I agree with Microsoft/Google/KDE's order. The author's situation is extremely rare, and the situation where someone wants "10" to be before "9" is far more common. Moreover, desktops don't label this sorting "alphabetical" (E: and it would really be "lexicographic"*), they label it "by name" (an informal criteria), so technically they're not lying.
> I miss the time when computers did what you told them to, instead of trying to read your mind.
You may be looking at that time through rose-tinted glasses. I don't like when computers lie to me either, but "mind-reading" is really helpful in ways we take for granted, like autosave. Desktops can have an option to sort files truly alphabetically, but the more common case should always be the default; that's the definition of "intuitive".
I'll add that I'm plenty "smart" enough to understand that "10" comes before "9" in a strictly alphabetical sense, and I still want my file managers to sort "9" before "10".
I don't want to put leading 0s before every single digit number in my file names. (And then potentially go back later and add another leading zero once something reaches three digits.)
---
All of my audiobooks are named in the format "Chapter 01.mp3", or "Chapter 001.mp3" in the case of books with more than 99 chapters. I do this because sometimes I put audiobooks onto music players that are too dumb to sort numbers properly, and adding leading zeros works everywhere.
I'm not sure I agree. I think I could be convinced if there was a unique and universal representation for numeric values using characters.
But we have so many textual representations of numeric values that I'm assuming the "mind-reading" goodness only works for a small subset. And the subset will be somewhat intuitive for developers but unlikely to be so for non-technical people.
For example, does the order handle numbers with fractions (decimal points)? If yes, does it require a at least one leading digit (zero)? Does a.12345 come before or after a.345?
Does it handle thousand separators? What about international thousand and decimal separators (e.g. Euro-style . for thousand separation and , for decimal separation).
Does it handle scientific notation?
If the answer is no to any of these questions, it's likely to lead to surprise/confusion.
It's like a feature request that initially sounds reasonable and useful but once you explore the requirements in detail you realize there are too many edge cases to be able to meet the request in a non-brittle way.
The sort rules are simple (1). Treat any consecutive sequence of digits as a number when sorting. So for example version numbers (which must be massively more common than decimals in filenames) work correctly, and 5.9 is indeed smaller than 5.10 and the latter is not identical to 5.1 .
Given that this idea goes back more than two decades, has been the default behaviour of the most used OSes for many years, with no major outcry, I think empirically we can be fairly certain that it does not routinely lead to a lot of surprises and confusion.
In considering the simplicity of the rule, I think you're using a developers perspective here where we automatically classify numbers and have a clear mental model of the separation between value and representation.
But I'm not sure how simple it would be to explain to a non-technical user why size_5, size_10 and size_15 are in order but size_0.25, size_0.5 and size_0.75 are out-of-order.
> with no major outcry
I'm regularly amazed at how little non-developer/technical users complain about strange and confusing behavior.
There's a user expectation that photo20.jpg comes after photo3.jpg.
There's no user expectation around whether photo1.jpg or photo01.jpg comes first. Just like there's no user expectation around whether photo1.jpg or Photo1.jpg comes first. Users also don't have the slightest idea about what order punctuation gets sorted in.
Just sort the things that matter in the way users expect (natural sort order) and come up with something reasonably consistent for the rest.
Here's a different scenario: filenames with dates in them. Consider September Budget and October Budget. September is the equivalent of 9, October of 10. Which comes first for natural sorting? Remember, the file modify date may not be useful here since you may have wrapped up the September budget on October 1st while the prior edit to the October budget may have been on September 20th.
The problem is that there is no such thing as natural, and it is quite hard to determine what is more common. (Quite often more common is culturally dependent or, worse, contex dependent).
>September is the equivalent of 9, October of 10. Which comes first for natural sorting? Remember, the file modify date may not be useful here since you may have wrapped up the September budget on October 1st while the prior edit to the October budget may have been on September 20th. The problem is that there is no such thing as natural
Yeah, but there is such a thing as "give a predictable and consistent way I can name the files so that they sort as I want everywhere" which (if different OSes don't try to be "smart") would have been to prefix them with the numeric date zero padded.
Given that 09 and 10 refer to months, that wont ever gonna be a problem. And if you want to differentiate them years too, you can prefix with 2025- or put them in a 2025/, 2026/ etc folder.
> It’s been about two thousand years since the number of months in a year has been increased.
What? What are you thinking of? The number of months in a year is always 12 or 13 in any calendar system because they start by reflecting the moon. If you mean the Christian calendar, it was fixed at 12 months to the year well over 2000 years ago. If you mean any calendar, it's probably been more like one year since the number of months in a year has been increased. 12 lunar months falls short of a solar year by about 11 days, so any given lunar calendar will generate an extra month about every three years, and there are lots of different lunar calendars.
(For example, the Chinese calendar occasionally repeats full months in order to keep the month of the year lined up with the season. Whenever this happens, there will be 13 months in the year, of which two share the same name.)
Are filenames likely to include those representations? I feel like probably not (can you even include commas in Windows filenames?)
More to the point of the article--if you want things sorted by date, sort by date. I think most laypeople aren't looking at long CHAR1234_5678 filenames anyway, they're looking at thumbnails and dates.
The most common date format used in Europe uses period separators so can often appear in filenames. Commas are probably more rare. Things like versions are often fractional like v1.3 or v1.11 and can appear embedded in filenames.
> can you even include commas in Windows filenames?
Yes.
> Use any character in the current code page for a name, including Unicode characters and characters in the extended character set (128–255), except for the following:
The following reserved characters:
The entire idea that numbers would be treated on a character by character basis rather than as numbers is somewhat intuitive for developers and not for non-technical people.
The answer to all of those questions is no for lexicographic ordering. Lexicographic ordering leads to surprise and confusion as a result.
> It's like a feature request that initially sounds reasonable and useful but once you explore the requirements in detail you realize there are too many edge cases to be able to meet the request in a non-brittle way.
It's been on windows and macOS for coming up on 25 years, and is in practically every modern UI. It’s reasonable.
Agreed.What's more, the idea that people learn to put leading zeros is wrong and impractical, unless you know in advance how many digits you need. When you go from version 5.9.17 to 5.10.0 you don't go back and relabel every existing folder as 5.09.17.
The today standard way of sorting is well defined, unambiguous, and natural. Lexographic has its place, but user facing interfaces ain't it.
Lexicographic order is great when you need an unambiguous criterion that will work the same in every implementation; but you only need that for automated processing, i.e. for coding.
For user-facing presentation, having 5.9.xxx before 5.10.xxx is simpler; the corner case that baffles users is having 5.1 and 5.10 before 5.2.
> I agree with Microsoft/Google/KDE's order. The author's situation is extremely rare...
Even if that were a valid reason for making it the default behavior, the real issue is they don't even give you the option to have the lexically correct sort order. They just decided to give you something that's not accurate and that's all you get.
A trend which is frustratingly, increasingly common.
It's trivial to allow customization behind menus. But we rarely get that anymore. Especially for sandboxes devices like phones.
It's a giant middle finger to users who want to actually use their devices as a tool, instead of simply a portal for more sales and marketing.
I think the only problem is that it's a surprise and mystery, particularly because "dumb" alphabetical sort has existed forever. When they "fixed this" for the 99% of regular users cases, they should have made it as separate "smart natural sort" option separate from the "strict alphabetical sort" option (next to date, size, etc). Simple and obvious, rather than surprisingly different from the decades of experience that even non-technical users already have.
It's not just the one decision though; there are literally thousands, maybe tens of thousands, of these decisions in most software. You want every single one of them to have an option? You want it to support every single combination? At some point, it is ridiculous. Sometimes you just have to decide how your software is going to work and not leave every single decision to the user.
You don’t let every decision to the user, you make good defaults, but leave the option to override to the user! And thousands isn’t scary as long as groups/tags/search work, so what’s ridiculous about empowering the user?
Increasing the number of different possible combinations of settings your software can be running with by a factor of one nonillion is not a choice I’d make if I wanted to have any confidence in its reliability and security.
How the files sort seems kinda important. It gets at the core behavior of the program. It's not something superficial like a default icon, which the user probably can change.
In a file manager? Any more than the displayed thumbnails, icon size, whether folders are separated from files, whether images are separated from videos, what video types are supported, what file types are opened inline, what the click and double click behaviours are, etc?
And yeah kde has settings for all these but kde is also known for being too configurable.
I might be wrong on this, but I vaguely recall that on macOS back when you could commonly option-click to reveal advanced options, if you held option when clicking a sort it would change how it sorted from alphabetical to lexical or vice versa. I’m not a thousand percent sure of it, though, I think when I needed it I was able to set a directory preference via terminal to change how a specific directory was sorted and it was an option there. MacOS had (or has) a lot of buried options which I presume date back to its origins as a Unix as well as a convenience to its developers. A lot of the command line utilities were hacked calls to graphical settings code though, so it wasn’t very stable version to version as the UI calls changed and nobody prioritized non-UI bug fixes or breaking changes. These days CLI is nearly forgotten or assumed to be an exploit vector - see Screen Time data for example.
But the alternative would be a surprise to people who assume "by name" will order numbers, including those who are new to technology (and I think most non-technical people who sort things manually unknowingly order numbers).
We want to minimize surprises and mysteries, but computers have so much hidden complexity it's impossible to eliminate them. If users were shown a full description of how every feature on their computer worked before using it, they'd quickly start ignoring the descriptions. There should probably be a tooltip or "manual entry" for "by name" for those who are curious, and it should never be labeled "alphabetical" because it's not. But cases like the author's, where he assumes a feature works differently than most people (including the designers) assume, can't be helped.
> and the situation where someone wants "10" to be before "9" is far more common.
I guess you mean "after"? Otherwise it seems to me you're agreeing with OP.
> desktops don't label this sorting "alphabetical" (E: and it would really be "lexicographic"*), they label it "by name" (an informal criteria), so technically they're not lying.
FYI the more formal name for the "by name" order is "natural sort order".
It’s more confusing. I thought the article was correct when they said -10 coming before -9. Why? Because they were talking about the strict alphabetical sort. They are already prepending zeroes to force the comparison to be 10 vs 09. So, yes, they were talking about ascending order, but not natural ascending order, but ascii sorting order where 10 is before 9 because the comparison isn’t 9 vs 10, but 1 vs 9.
It was only clear to me because I could guess where they were going. They were complaining about natural sort vs alphabetical sort, which is a case I’ve run into many times, so I could see the argument coming.
The irony to me was that they were already altering how they named files to fit what they thought the computer wanted by prepending a zero to get a proper alphabetic sort. And even after that, some computers didn’t follow their idea of what it should be doing.
People sorting their files in alphabetical order but who want numerical values in their files to be sorted digit by digit instead of as numbers is the rare case.
I might go further in my ideal sorting algo which would be normalize capitalization and ignore all alphanumeric characters and treat them all as separators.
I have some beef with microsoft, that you can only change this at the Computer level, not per user (see registry key below). Also they call it natural sorting for users, but logical sorting internaly. Unify your termini!
TIL they are called "hives". Windows Registry is an interesting thing. Even casual users have to interactive with it once or twice w/o fully understand it.
Raymond Chen explained why a registry file is called a “hive”:
Because one of the original developers of Windows NT hated bees. So the developer who was responsible for the registry snuck in as many bee references as he could. A registry file is called a “hive”, and registry data are stored in “cells”, which is what honeycombs are made of.
The file sorting isn’t something relegated to niche users because of the prevalence of tv episode file name sorting (eg S01e01) and it has necessitated the leading zeroes to make it work properly with “alphabetical sorting”.
The author wants the "worse" sort, one based on ASCII/Unicode codepoints, without any intelligence for numbers that 99% of GUI users want.
For their purposes, they've assumed something about the implementation, to the point that a convenience feature is actually a misfeature for them. But the author here is probably a developer, or close to one, so they do not represent the needs of most people using computers.
Understanding the target audience for your product results in very different design decisions. Better is better might be great for products, but worse is better is probably better for systems that need to grow and evolve.
It's an issue of mental models. As a developer, his mental model is one of how naive software would sort items with mixed numbers in them. Most people, of course, naturally sort 10 after 9 -- their mental model doesn't contain software developer assumptions.
> The author wants the "worse" sort, one based on ASCII/Unicode codepoints, without any intelligence for numbers that 99% of GUI users want.
I want the author's opinion on how caplital and lowercase letters should be sorted. Do they follow strict ASCII/Unicode codepoints, or do they normalize into actual alphabetical order and sort upper/lower within each letter?
> I want the author's opinion on how caplital and lowercase letters should be sorted. Do they follow strict ASCII/Unicode codepoints, or do they normalize into actual alphabetical order and sort upper/lower within each letter?
I prefer the strict ASCII / Unicode sorting (all capitals first, then all lowercase).
This feels like the right moment to mention "ch", which is considered a letter in orthodox Czech, sorted between "h" and "i". The problem is, you can't reliably distinguish between "ch"-the-letter and "ch" as just "c" and "h" combined, which are present in loan words but also some original Czech compound words.
So if you're doing it "properly", sorting strings in Czech involves understanding the etymology of every word.
>Understanding the target audience for your product results in very different design decisions
This is an excuse. Just add an option to sort both ways. It isn't hard.
There is no target audience in this planet that benefits from less options or less features. Even if you had the features under an "advanced mode" UI that's still a better software than not having the feature in first place.
Have people forgotten the 80/20 rule? Most features will be used by only a small slice of users, that doesn't mean they're out of scope.
Sorry, I'm just kind of exhausted of software not being able to do the most obvious things because it didn't align to some perfect vision of how the user should be.
The most irritating circumstance for this is looking for files named with a hash:
3ea4f...
...
97dce...
...
126b9...
This is one of the settings I immediately turn off on Windows via the registry key mentioned in the other comments here.
I miss the time when computers did what you told them to, instead of trying to read your mind.
These days, it's more like "trying to change your mind". I absolutely hate the "the user is wrong" authoritarian mentality that unfortunately has infected a ton of software, even open-source.
Exactly. This is even more annoying when it isn't exactly a hash, but some gibberish you cannot really make sense of, which does have a numeric section in them: like a user ID, or unix time, or who knows what else it could be, but you are trying to visually find a file abcd89764237 somewhere after abcd683426834, and it isn't evident why you cannot, unit you notice that the latter has more digits in its "ID" for some reason.
I almost always want the version-sorting that's being presented in this article, rather than an "alphabetical" sort. But on the other hand, it absolutely seems like a valid bug that this is presented as an "alphabetical" sort, rather than something like "alphabetic/numeric" or similar. In other words, a problem of labeling rather than one of sorting.
It’s not being presented as an alphabetical sort, though. The author assumed that sorting by name meant an alphabetical sort, but that’s not how it’s labeled.
Author here - I Agree both with you and with the parent's comment. Having two options in the "sort by menu" - like "Name (natural)" and "Name (strict)" or something - would have solved everything.
> The problem is imposing it on the user with no warning or option to turn it off.
You can say that about every single design decision made about every product.
The gripe about this particular feature seems misplaced because almost all users will want the sort that's offered and the actual alphabetical sort is likely the desire of a more advanced user who, in fact, is offered a choice through registry editing and/or using a more advanced cli option for the occasion they might need an alternative sort.
Obviously. All I'm saying is that this particular decision ought not to have been taken from the user. Real alphabetical order is not an unreasonable thing to want.
“Real alphabetical ordering” is incredibly nonspecific. It’s underspecified even for ASCII-US, but essentially meaningless for those of us in 2025 who need to handle Unicode.
How do capital letters sort relative to lowercase letters? How do letters sort relative to digits? How do you consider code points that can correspond to different letters in different lettering systems with different ordering? How do you handle diacritics? Do you want the behaviour to be stable through Unicode normalization? Should it differ based on the character encoding? Should different representations of the same character, such as blackboard lettering or circled numbers, be sorted with other representations of the same character or grouped separately?
You can come up answers for these questions, but there’s no unambiguously correct option. The least subjective option is sorting based on encoded byte representation (if that is even specified), but that is not “alphabetical” and would not be intuitive to most users.
Notably, some versions of “sort” on Linux have version sort nowadays. sort -V
I actually don’t know exactly how it works internally and it is a little bit magical, but I use it all the time when looking through my files because it just sorta works in most cases. Of course a nice thing about it is easy to turn on or off.
“There are additional complications in certain languages, where the comparison is context sensitive and depends on more than just single characters compared directly against one another,
[…]
Numbers. A customization may be desired to allow sorting numbers in numeric order. If strings including numbers are merely sorted alphabetically, the string “A-10” comes before the string “A-2”, which is often not desired. This behavior can be customized, but it is complicated by ambiguities in recognizing numbers within strings (because they may be formatted according to different language conventions). Once each number is recognized, it can be preprocessed to convert it into a format that allows for correct numeric sorting, such as a textual version of the IEEE numeric format.”*
I think those file browsers made the right choice, even given that they don’t (as in this example) always do the right thing.
I thought this was pretty well known. E.g. the macOS Foundation library even exposes NSString.localizedStandardCompare() [1] which implements the sorting algorithm used by Finder, and should be used by any well-behaved macOS application. Windows uses StrCompareLogical [2].
I would have assumed it worked the same as ls, so I found the article interesting. But now that I know, I think this way is better.
I can’t think of any case where I would need purely alphabetical sort. In most photo browsing apps, photos will be sorted by timestamp rather than filename. If I really needed it to sort properly in file explorer, I would try sorting on created date. And failing that I would probably just normalize the file names.
> Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means. So when you ask them to sort your files alphabetically, they don’t. Instead, they decide that if some piece of the file name is a number, the real numerical value must be used.
Well, no. You don't actually ask them to sort in alphabetical order. You ask them to sort "by name", and that is up to their interpretation. And they choose the interpretation that (per their reasoning, and possibly some actual data) seems most likely to correspond to what the user wants.
Maybe future versions of those OSes will add a rule that says that if any of the number groups have leading zeros then it reverts back to actual alphabetic order. Or maybe they'll give you configurable options. (Maybe some of them already do.)
Clearly a leading zero means the number is in octal (but only if all the subsequent digits are between 0 and 7). I think that would lead to the most intuitive results.
> And they choose the interpretation that (per their reasoning, and possibly some actual data) seems most likely to correspond to what the user wants.
Yes, that make sense, but the problem is that this interpretation changed in the last 10 (15? 20?) years. It used to be that "by name" meant "by name, il alphabetical / lexicographical order" in pretty much every file manager.
Sorting so "foo9" is before "foo10" is called natural sort. I found out about natural sort a week ago and I am thrilled that my programs now print their output in a sensible order. Give natural sort a try and see if it improves your life too :-)
for i in $(seq 2 10) ; do
touch img_$i-hn.txt
done
ls img_* | sort -V
img_2-hn.txt
img_3-hn.txt
img_4-hn.txt
img_5-hn.txt
img_6-hn.txt
img_7-hn.txt
img_8-hn.txt
img_9-hn.txt
img_10-hn.txt
And we have "sort -h" to sort the output of e.g. "du -sh *" properly.
"I created the Alphanum Algorithm to solve this problem. The Alphanum Algorithm sorts strings containing a mix of letters and numbers. Given strings of mixed characters and numbers, it sorts the numbers in value order, while sorting the non-numbers in ASCII order. The end result is a natural sorting order."
There are many older instances of that, such as "versionsort" from various Linux tools and libraries. I think this has likely been independently recreated several times, with various subtle differences.
I get it, but if all these major operating systems are handling this same ambiguous [0] situation in the same way, perhaps one needs to reevaluate their mental model or expectations.
Am I out of touch? No, it's the operating systems who are wrong
ls sorts filenames strictly lexicographically, comparing character by character, so e.g. "055436307" is compared as the characters "0", "5", "5", etc. so it sorts before "121134" because "0" is less than "1". if all compared characters match and one string ends, the shorter one comes first. Symbols like _ are just more characters, and their position relative to digits depends on the locale’s collation table.
Google Drive uses ICU collation with the numeric option enabled, which treats each consecutive sequence of digits inside the filename as an integer. so "055436307" is parsed as the number 55,436,307, while "121134" is parsed as 121,134. and since 121,134 < 55,436,307 then "121134..." comes before "055436307..." even though lexicographic order would suggest the opposite. and i think when two digit runs have the same numeric value, the shorter run comes first; if runs are equal and the string continues, then normal character comparison resumes, including any underscores or suffixes
Earlier this year I submitted a bug to VSCode about sorting Playwright tests in the alphabetical-and-numerical order that VSCode favours, after Playwright told me it was a VSCode issue.
Some people rushed to fix this as I'd done some diving into the issue and presented the relevant information and code, so now VSCode's Playwright test list uses the same sorting mechanism as the rest of VSCode.
Sadly, the underlying Playwright does not receive that order from VSCode so it still actually runs sequentially-numbered tests in strict alphabetical order. :(
Well, lots of interfaces don’t say “alphabetical” anymore, they say “name” or some variant, and then they can define it however they want, regardless or because of the frustration it causes users but not some other users which will now be inverted for long term-frustration averaged user experience.
Nah that's not just you. That is an unnatural way to sort things because that's not how numbers are ordered. I remember when Windows changed to sorting numbers by their value and, despite my programmer brain finding it strange in a way, I was super happy to have files display in an order that actually made sense.
I felt a little bad about this snark but actually, author barely understands their own use case (says they want alphabetical order but they actually want something more) and barely understands the UI they're using (says they asked for alphabetical order but none of the file managers they used says it has any such setting) and then they go on to claim this is to satisfy dumb users:
> Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means.
The so-called "natural" sort makes sense for version numbers and enumeration (without zero-padding) but I'm more often dealing with file names with a datetime (like in the article), a hexadecimal hash, or just randomized string of characters that includes numbers. In those cases "natural" sort makes it harder to find the file you're looking for.
Even when files are enumerated it's pretty rare to have more than 9 parts and no zero-padding, whereas there are almost always multiple consecutive digits in the use cases for which "natural" sort is not a good fit for. It just feels like a bad default, at least for a programmer's workload.
To answer the question in the article, I’m pretty sure Windows Explorer (and probably File Manager before that) has sorted filenames this way for at least 30 years.
I got used to naming files/folders with leading zeros when I want them to be sorted alphabetically (for example payslips/invoices, etc).
But I'm a tech guy, I know what does "alphabetically" mean in the tech world. And it probably is not what common folks mean when they think "alphabetically" outside the tech world.
Edit: in fact, if I recall correctly, the proper term for this kind of sort (the one OP wants) is alphanumeric sort.
I also got used to it, but especially when writing short scripts that generate numbered files it gets annoying to have to pad with zeroes every time, and also precommit to a specific amount of digits you want to allow (finding a compromise between adding a ridiculous amount, like 20, and using only 4 despite knowing the script might one day surpass 10⁵ files).
The natural numbers are ordered. Let me use its ordering instead of having to rely on an ad-hoc lexicographic fixed-length tuple representation of decimal digits, without any padding. My position is that numbers in filenames should always be considered atomically unless explicitly instructed otherwise.
If there were no issues of backwards compatibility, I would thus advocate for changing ls. Eza (maintained fork of Exa, Rust-based ls alternative) actually does sort this way by default, much to my delight.
There's a Group Policy setting in Windows: Computer Configuration\Administrative Templates\Windows Components\File Explorer\Turn off numerical sorting in File Explorer
Group Policy has so many essential settings I hurry to change with every isntall. I wish Windows would expose more of them to the user in ordinary settings.
I think if we (in our industry in general) had REAL agile, and not pseudo-waterfall "the designers design it, the engineers implement it, QA QA's it, and then we lay everyone off because they're no longer needed" (but, loophole alert! We did daily standups and used Jira, so it was "agile" the whole time!), then we'd have a snowball's chance in hell of actually having a reasonable solution to this. Off the top of my head, this seems like something that should be a setting in control panel. But, because everyone assumes (contra to agile) that the designers "got it right the first time", this kind of improvement can't happen.
For some inexplicable reason, Plex just throws its hand up on non-ASCII characters and puts them first.
In Norway we have three extra letters, æøå, and they're at the end of the alphabet after z. But in Plex, I have Øystein Sunde[1] placed before any other in my music library.
Now in the 1990s I would forgive US software for such a thing, but it's 2025...
If only we could represent sort order by some structural form of decision logic which also embeds encoding a regular .. well.. expression matching a pattern..
I think the real issue here is that two Android phones take photos with incompatible naming schemes.
I am sure that at some point someone thought the milliseconds should or should not be separated from the seconds and made that change without thinking through the consequences.
I was joking. Really I would sort file names lexicographically. But the way Debian sorts version numbers is interesting and seems like a good way of handling that particular situation.
I find the term "natural" to be inadequate here, there is nothing natural about sorting strings in this particular fashion compared to another. It should be given a more descriptive name, like "Number-aware alphabetic" or something like that as to actually give a hint about what it does.
Heh. One of the bugs that once caused me to bang my head against the wall was caused by the Estonian language. Its alphabet has Z following S and Š. So the "foolproof" regexp to match the letters '[a-za-Z]' was misfiring for some entries.
Be glad you don't have to deal with non-ASCII characters: acute/grave/tilde/umlaut/diaresis/etc. accented characters, dotted vs. dotless 'i' (Turkish), barred i (an 'i' with a sort of dash through the middle, used in some languages for a sort of schwa-like vowel), thorn, not to mention non-Roman characters. And different languages sort the same characters differently, so you can't just pay attention to their Unicode values. (@cubefox has a post here pointing to the Unicode Consortium's doc about sorting)
Renaming things to make them queue correctly (I usually couldn't care less about visual sorting, I use a terminal) is by far my #1 task by LoC and frequency of occurence, and by far the most annoying. Metadata can be very helpful to obviate this issue, but it usually just leads to another problem where you now need metadata editors and readers in addition to the "user-visible" name metadata. It's frustrating.
I think the algorithm is probably incorrect. A number starting with 0 should be treated lexically not numerically. Otherwise you have a situation where img_1_01.jpg and img_01_1.jpg does not have a complete ordering.
> Otherwise you have a situation where img_1_01.jpg and img_01_1.jpg does not have a complete ordering.
(Good) "natural sort" implementations generally have ways of handling ties like this. It's similar to the problem of case-insensitive sort over case sensitive sets.
I'm often yelling at the software on my computer: "STOP TRYING TO HELP ME!"
It's like having a toddler help you make a meal. It wants to be involved and recognized so badly. Meanwhile I'm starving and just want to get the food done as quickly as is possible and I'm constantly tripping over this little ball of misguided efforts.
Please. Stop trying to be smarter than me. You often can't, and when you get it wrong, you make it measurably worse. If you insist on doing this please give me the "Expert Mode" setting back so I can flatly disable ALL OF IT with one click.
If you don't like the default natural sorting order, you can just change it in Dolphin. Settings > Configure Dolphin > View > Content Display > select anything other than "natural". You can even pick if you want case sensitivity or not.
The OS doesn't think you're too stupid to understand sorting, it relies on you being smart enough to figure out where the setting is located. In this case, four levels deep is probably too much to ask from users if they will write an entire blog post like this before finding the toggle.
I thought this was going to be a deep dive into what "alphabetical" means and how that's itself not a universal term between locales, what with so many different collation preferences.
That would likely have been a more useful article for the average developer. It is extremely hard to be aware of all the ways strings of different locales can defy our intuitions.
I agree that the base functionality of just sorting character by character can be occasionally useful. However I would really be interested in seeing why you believe this to be the correct choice for user-facing graphical file managers, as its evident problems with typical usage seem more salient compared to the edge cases as illustrated in the article.
When you get a bunch of files (let's say 1000+) without leading zeroes, this is a blessing. But I get the author's frustration, the expected behavior is not there, instead, he gets magical sorting that is wrong for his use case. I'm not sure what the ux should be, and maybe the algorithm here could be smarter, but it's a trade-off.
The expected behaviour is ambiguous (and thus subjective). Older versions of windows shipped with alpha sort. New versions ship with natural alpha sort. According to the UX designers over at Microsoft (and surely the user feedback), natural sort _is_ the expected behaviour.
I certainly agree with natural sort being the expected behaviour too.
>Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means.
i really really hate this framing, and i see it far too often. no, the operating system developers did not make a value judgement about their users. they observed their users to find out what behaviour was expected, and they designed the behaviour of the system to match the behaviour that the majority of users expect.
and then you made an incorrect assumption about how the system works, and decided that your incorrect assumption means everybody else is dumb and you're the only smart person in this situation?
Even if you are a file naming Einstein and you always zero pad your integers to exactly right length, we have this thing called the internet, where you can download other people's files.
This is one of the big ways that LLMs are going to change the game for UX. Your operating system is going to have some sort of 'butler', which knows all of your preferences, and the butler will go through the APIs and man files and informational dialogs of every app you use and auto-configure them.
Then if you want something to change, just ask the butler. If the app is open source and doesn't support the requested feature, the butler might even be able to code it up.
This was a fun thing to realize in my early days of programming in Delphi. I guess the author will soon realize why old systems name things ticket "00001" and so on.
I've encountered a tangential problem to this with package versioning on Linux distros. Thankfully it was not too hard to write an algorithm to compare versions (thanks AI!).
Another problem which annoys me to no end is that most file managers and file selection boxes put directories before files.
This makes it hard to find the file that was most recently changed, for example. Which is an action that is extremely common. (In fact, why does my file manager not have a most-recently-used shortcut?)
I rename all of my photos upon import using the created date, formatted as `YYYY-MM-DD kk:mm:ss`.
But it would frankly be great if most file browsers just let me sort photos based on metadata. But then I just end up in a dedicated photo browser, instead.
Call lexicographic order "sort by name" as it's called now, and call dumb character-by-character sort "plain" or something like that. I'm not a designer, maybe there are more intuitive names, but come on. This isn't an intractable problem.
I have the same problem on Nemo. More specifically, I had made a small app that displayed files of a directory in alphabetical order, and then when I look at it in Nemo it isn't the same order because I didn't implement their smart algorithm.
I fail to see how this is a "problem"? You implemented a sorting mechanism that was useful to your application, while Nemo implemented another which as this thread demonstrates seems to be much more useful and intuitive for the average user. This is also of course not specific to Nemo, as no 'modern' file manager on Linux sorts filenames like it's 1980 and all you are able to feasibly do is step through the bytes.
> But 1 is smaller than 9, so file-10.txt should be first in alphabetical order. Everyone understands that, and soon people learn to put enough leading zeros if they want their files to stay sorted the way they like.
No. Not “everyone understands that”. Natural sort happens in real life and everyone understands that. Only those who understand ASCII — not the average user of graphical file managers — will deduce the reason for your definition of “alphabetical order”.
> Now that I know what the issue is, I can solve it by renaming the files with a consistent scheme.
In Total Commander, there is a function in the options to sort strict by numerical char code. It will sort those files correctly. Unfortunately, it will also sort "10.txt" before "2.txt".
---
In all file managers, I miss an API point where one can give a userdefined sorting function for the file and folder list.
What do you mean by "Unfortunately"? This appears to be the only correct conclusion from the algorithm you selected, you can't eat the cake and have it too.
Regarding your second point, that's not really what a graphical file manager is for, I think. At this point (likely even earlier) you would be better off just writing a simple script in the scripting language of your choice. (If going for something fancy, you could also implement a FUSE based on symlinks for the original files, where the filename is prepended by a sort key. This would work for every major file manager and you could manipulate the files in mostly the same way as before.)
Paragraph 1: I speak of a sorting method which splits the filename at the boundaries between numbers and non-numbers, and sorts by the parts of the resulting tuple, the numbers naturally (10 comes after 2) and the rest by numerical char code.
Paragraph 2: I am not sure what you mean here with writing a script. The graphical file manager shall sort its file list using the sorting function I hand over to it.
"That's not really what a graphical file manager is for". Says who? Every software which has a plugin system does that, why should a file manager not?
Sorting by name (collation) is waaay tricker than simply figuring out how to parse the numbers.
The International Components for Unicode library implements the Unicode Collation Algorithm, which depends on the language code and region of the locale, and looks up the quirks for each locale in the Common Locale Data Repository.
It's a much better idea to just use the standard ICU library or platform specific libraries (which are often build on ICU like JavaScript's Intl.Collator), instead of trying to hot dog it by rolling your own.
>ICU provides the following services: Unicode text handling, full character properties, and character set conversions; Unicode regular expressions; full Unicode sets; character, word, and line boundaries; language-sensitive collation and searching; normalization, upper and lowercase conversion, and script transliterations; comprehensive locale data and resource bundle architecture via the Common Locale Data Repository (CLDR); multiple calendars and time zones; and rule-based formatting and parsing of dates, times, numbers, currencies, and messages.
>The Unicode collation algorithm (UCA) is an algorithm defined in Unicode Technical Report #10, which is a customizable method to produce binary keys from strings representing text in any writing system and language that can be represented with Unicode. These keys can then be efficiently compared byte by byte in order to collate or sort them according to the rules of the language, with options for ignoring case, accents, etc.[1]
>Unicode Technical Report #10 also specifies the Default Unicode Collation Element Table (DUCET). This data file specifies a default collation ordering. The DUCET is customizable for different languages,[1][2] and some such customizations can be found in the Unicode Common Locale Data Repository (CLDR).[3]
>The Common Locale Data Repository (CLDR) is a project of the Unicode Consortium to provide locale data in XML format for use in computer applications. CLDR contains locale-specific information that an operating system will typically provide to applications. CLDR is written in the Locale Data Markup Language (LDML).
>Among the types of data that CLDR includes are the following:
Translations for language names
Translations for territory and country names
Translations for currency names, including singular/plural modifications
Translations for weekday, month, era, period of day, in full and abbreviated forms
Translations for time zones and example cities (or similar) for time zones
Translations for calendar fields
Patterns for formatting/parsing dates or times of day
Exemplar sets of characters used for writing the language
Patterns for formatting/parsing numbers
Rules for language-adapted collation
Rules for spelling out numbers as words
Rules for formatting numbers in traditional numeral systems (such as Roman and Armenian numerals)
Rules for transliteration between scripts, much of it based on BGN/PCGN romanization
Tricky collation examples:
sv-SE (Swedish): å, ä, ö are separate letters at the end of the alphabet, not variants of a or o.
de-DE (German): ä, ö, ü may sort as ae, oe, ue in some contexts, or as distinct letters. ß sometimes sorts as ss.
tr-TR (Turkish): dotted i (i) and dotless ı are different letters; I sorts with ı, not with i.
es-ES (Spanish): traditionally ch and ll were treated as single letters with their own place in the alphabet.
cs-CZ (Czech): ch still counts as a unique letter, sorted after h.
da-DK / no-NO (Danish/Norwegian): ø comes after z.
is-IS (Icelandic): þ (“thorn”) is part of the alphabet, after z.
fr-FR (French): accents usually ignored in sorting, so é = e, but not always depending on collation settings.
el-GR (Modern Greek): tonos accents, final sigma ς vs. σ, etc.
nl-NL (Dutch): the digraph “ij” is often treated as a single letter, and capitalized as “IJ”. In dictionaries and phone books it often sorts as a single letter under “I”, but sometimes is listed after “X” depending on tradition.
Then you get into non-Latin languages like, Chinese, Japanese, and Korean collation, which gets hairy with radicals, kana order, and stroke count.
Also different locales have different ways of representing numbers, like switching between "," and "." as separators and decimal points.
ICU supports integer only "natural" numeric collation, so anything more complicated like versions, floating point, negative numbers, hex, thousands separators, fractions, roman numerals, etc, you'd have to build on top of ICU.
ICU doesn't support incomprehensible dead languages like Latin or Ancient Greek (it does however support French ;). It does support Roman numeral formatting, but not collation, which would be pretty tricky and ambiguous.
A nuanced but common example that ICU/UCA/CLDR helps with is a menu to select the current locale: you have to translate each language's name into the current locale, and also sort them in the current locale. On top of different collations they can also have totally different spellings, like "United States of America" is "Verenigde Staten van Amerika" in Dutch. This makes it challenging for users to find their own language when the locale is set wrong! You just can't win.
Not to mention emojis! Which comes first: The chicken or the egg? The taco or the poop?
Also, the Mac Finder switches ":" and "/" for historical reasons (HFS used to use ":" as a directory separator instead of "/"), so you can create a file name like "9/11 Attack" in the Finder, which actually gets the underlying Unix filename "9:11 Attack". Don't believe me? Rename a file in the Finder to include a slash, which you know is impossible to represent as a Unix file name. Then go "ls" the directory in the shell.
The Mac Finder weirdly collates "/" after "9" because under the hood it’s really storing it as ":", which sorts before "0". But it also has other punctuation collating inconsistencies, sorting "," and ";" and others after "0" too. Definitely not ASCII order -- I'm not sure what rules it uses, but it's different than "ls".
However, while it's generally true you can't have "/" in Unix file names, NFS used to trustingly let clients rename Unix files to include a "/" in their name, which the Gator Box AppleTalk/Ethernet gateway let you do with the Mac Finder (pre OS/X), which would silently corrupt your "dump" backups on the Unix NFS server, so you would not learn about it until you tried to retrieve your files and "restore" crashed.
>Another reason that NFS sucks: Anyone remember the Gator Box? It enabled you to trick NFS into putting slashes into the names of files and directories, which seemed to work at the time, but came back to totally fuck you later when you tried to restore a dump of your file system.
>The NFS protocol itself didn't disallow slashes in file names, so the NFS server would accept them without question from any client, silently corrupting the file system without any warning. Thanks, NFS!
By the way, there seems to be a "standard" way to sort strings:
> Unicode Technical Report #10 also specifies the Default Unicode Collation Element Table (DUCET). This data file specifies a default collation ordering.
I assume this mainly aims at giving a reasonable compromise between the different dictionary and phone book sorting rules of various languages (and even locales), which should give reasonable results for most languages. I assume this also puts "Alice2" before "Alice10".
> 1.9.2 Non-Goals
>
> The Default Unicode Collation Element Table (DUCET) explicitly does not provide for the following features:
> [ ... ]
> Numeric formatting: numbers composed of a string of digits or other numerics will not necessarily sort in numerical order.
> Of course, the user who named those files probably wants file-9.txt to come before file-10.txt. But 1 is smaller than 9, so file-10.txt should be first in alphabetical order. Everyone understands that, and soon people learn to put enough leading zeros if they want their files to stay sorted the way they like. Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means. So when you ask them to sort your files alphabetically, they don’t. Instead, they decide that if some piece of the file name is a number, the real numerical value must be used.
I think there are many things wrong with your assessment of the situation.
First, where does it say in these file managers that they're sorting by alphabetical order? I see that you've specified that you want the files sorted by name, but I don't see that you've specified you want them sorted by name alphabetically. And what does "alphabetical sort" even mean when you're sorting characters which are not letters? What you mean is probably "lexicographical sort".
Second, you admit yourself that users probably want natural sort. Why would you expect these products to do the thing which they know users usually don't want by default? That just seems like bad design to me. They know users usually want natural sort, and you know users usually want natural sort, so why would you expect the default behaviour to be a lexicographical sort?
Third, just like how you've learned to work around the lack of natural sort in poorly designed products of years past by adding leading zeroes, you can just add trailing zeroes to get the lexicographical ordering that you want. Why do you seem to be implying that the latter is more user-hostile than the former? It doesn't make sense to me. A decision had to be made about what sort to use and they picked the one that most people want. Isn't that what we should be expecting in a product that caters to its users?
I see in other comments you've suggested that there should be a separate option for choosing between lexicographical sort and natural sort. But in the past, when lexicographical sort was the only option, why weren't you complaining about it being user-hostile to only have one option then? Why is it only when the default is something you're personally not used to that it warrants complaint? And where do we stop, do we have separate controls for every single sortable string field to determine whether it should be sorted lexicographically or naturally? Or just the name field? Don't you think that is going to lead to interface bloat?
If I understand the article, the author wants magic :)
I take it to mean they want the system to know file_9.txt is less then file_10.txt.
I never saw that happen in any OS, so I do not know what he is referring to. Maybe whatever that old system was, it sorted by create time as opposed to file name.
So, the author can try and create "aisort" that will look at all file names and add leading zeros to the file numeric portion, sort, then remove the zeros added. That will probably as slow as s***t and use gobs pf memory, depending on the number of files.
Author here - My surprise stems exactly from the fact that for the last few years I have exclusively managed my files via a the UNIX shell, which behaves in the classical way.
When I started using Linux as my daily driver after many years of Windows (but with familiarity with UNIX systems going way back), I knew it would be like that in the terminal, but it still took some adjustment. But actually, Nemo does the same "natural sort" thing, and also sorts case-insensitively.
That's not what the author says- they said that file managers actually are somehow sorting file-9.txt before file-10.txt, and it's breaking real alphabetical ordering.
i think it's the opposite, that they _want_ file_10.txt to come before file_9.txt by default, but that file explorers fail at this. it's rare that i want true alphabetical sort, but it's convenient for cases like tfa where alphabetical sort is more predictable if i have filenames that look like <letters>_<numbers-of-same-length>.txt.
Digits and any other characters that are not A-Z and a-z should not get sorted. That's the true result of doing what you asked and not what you meant. Pedantic, but that's why we are here.
Not this keyboard not this chair, but the problem is with idiots between keyboards and chairs.
The author is not the ID10T it’s the other general users.
The author is intelligent enough to recognize that this is not alphabetical sort, but the term that they are looking for to describe the sort that they see in dolphin windows, google etc. is *lexical* sort, not alphabetical.
The engineering problem is ID10Tic not technical. How do you educate an illiterate public on what the difference between alphabetical and lexical sort is in practice?
You can’t, so you engineer around it and call lexical sort alphabetical.
I agree with Microsoft/Google/KDE's order. The author's situation is extremely rare, and the situation where someone wants "10" to be before "9" is far more common. Moreover, desktops don't label this sorting "alphabetical" (E: and it would really be "lexicographic"*), they label it "by name" (an informal criteria), so technically they're not lying.
> I miss the time when computers did what you told them to, instead of trying to read your mind.
You may be looking at that time through rose-tinted glasses. I don't like when computers lie to me either, but "mind-reading" is really helpful in ways we take for granted, like autosave. Desktops can have an option to sort files truly alphabetically, but the more common case should always be the default; that's the definition of "intuitive".
* https://news.ycombinator.com/item?id=45404022#45405279
I'll add that I'm plenty "smart" enough to understand that "10" comes before "9" in a strictly alphabetical sense, and I still want my file managers to sort "9" before "10".
I don't want to put leading 0s before every single digit number in my file names. (And then potentially go back later and add another leading zero once something reaches three digits.)
---
All of my audiobooks are named in the format "Chapter 01.mp3", or "Chapter 001.mp3" in the case of books with more than 99 chapters. I do this because sometimes I put audiobooks onto music players that are too dumb to sort numbers properly, and adding leading zeros works everywhere.
But it would be nice if I didn't have to do that.
I'm not sure I agree. I think I could be convinced if there was a unique and universal representation for numeric values using characters.
But we have so many textual representations of numeric values that I'm assuming the "mind-reading" goodness only works for a small subset. And the subset will be somewhat intuitive for developers but unlikely to be so for non-technical people.
For example, does the order handle numbers with fractions (decimal points)? If yes, does it require a at least one leading digit (zero)? Does a.12345 come before or after a.345?
Does it handle thousand separators? What about international thousand and decimal separators (e.g. Euro-style . for thousand separation and , for decimal separation).
Does it handle scientific notation?
If the answer is no to any of these questions, it's likely to lead to surprise/confusion.
It's like a feature request that initially sounds reasonable and useful but once you explore the requirements in detail you realize there are too many edge cases to be able to meet the request in a non-brittle way.
The sort rules are simple (1). Treat any consecutive sequence of digits as a number when sorting. So for example version numbers (which must be massively more common than decimals in filenames) work correctly, and 5.9 is indeed smaller than 5.10 and the latter is not identical to 5.1 .
Given that this idea goes back more than two decades, has been the default behaviour of the most used OSes for many years, with no major outcry, I think empirically we can be fairly certain that it does not routinely lead to a lot of surprises and confusion.
(1) https://en.m.wikipedia.org/wiki/Natural_sort_order
> The sort rules are simple
In considering the simplicity of the rule, I think you're using a developers perspective here where we automatically classify numbers and have a clear mental model of the separation between value and representation.
But I'm not sure how simple it would be to explain to a non-technical user why size_5, size_10 and size_15 are in order but size_0.25, size_0.5 and size_0.75 are out-of-order.
> with no major outcry
I'm regularly amazed at how little non-developer/technical users complain about strange and confusing behavior.
> Treat any consecutive sequence of digits as a number when sorting.
Based on this description, I have no idea how the following would be sorted:
• photo.jpg
• photo1.jpg
• photo01.jpg
• photos.jpg
Does it matter?
There's a user expectation that photo20.jpg comes after photo3.jpg.
There's no user expectation around whether photo1.jpg or photo01.jpg comes first. Just like there's no user expectation around whether photo1.jpg or Photo1.jpg comes first. Users also don't have the slightest idea about what order punctuation gets sorted in.
Just sort the things that matter in the way users expect (natural sort order) and come up with something reasonably consistent for the rest.
> There's a user expectation that photo20.jpg comes after photo3.jpg.
my user expectation is the opposite
i get what you're saying but it's not achievable in practice, at least not consistently
I just tried it on Mac, its sorted in the order you listed. Extending it a bit, the order is:
photo1 photo01 photo001 photo0001 photo2
So the shorter representation of the same number comes first. It does make intuitive sense to me.
Ah, the classic filenames with decimal points and scientific notation in them, so common...
Here's a different scenario: filenames with dates in them. Consider September Budget and October Budget. September is the equivalent of 9, October of 10. Which comes first for natural sorting? Remember, the file modify date may not be useful here since you may have wrapped up the September budget on October 1st while the prior edit to the October budget may have been on September 20th.
The problem is that there is no such thing as natural, and it is quite hard to determine what is more common. (Quite often more common is culturally dependent or, worse, contex dependent).
So the argument is that because this doesn't solve the challenge of ordering all possible strings by semantic meaning it should not be used?
Even though it increases the match between semantic meaning and string sorting in many important cases and is a simple and consistent rule?
>September is the equivalent of 9, October of 10. Which comes first for natural sorting? Remember, the file modify date may not be useful here since you may have wrapped up the September budget on October 1st while the prior edit to the October budget may have been on September 20th. The problem is that there is no such thing as natural
Yeah, but there is such a thing as "give a predictable and consistent way I can name the files so that they sort as I want everywhere" which (if different OSes don't try to be "smart") would have been to prefix them with the numeric date zero padded.
Date is already in the metadata, it doesn't need to be in the filename.
the one true way: budget_09.csv, budget_10.csv
Then `budget_100.csv` comes by and now you need to rename 99 files.
Given that 09 and 10 refer to months, that wont ever gonna be a problem. And if you want to differentiate them years too, you can prefix with 2025- or put them in a 2025/, 2026/ etc folder.
It’s been about two thousand years since the number of months in a year has been increased. I don’t think we’re getting 88 new ones anytime soon.
> It’s been about two thousand years since the number of months in a year has been increased.
What? What are you thinking of? The number of months in a year is always 12 or 13 in any calendar system because they start by reflecting the moon. If you mean the Christian calendar, it was fixed at 12 months to the year well over 2000 years ago. If you mean any calendar, it's probably been more like one year since the number of months in a year has been increased. 12 lunar months falls short of a solar year by about 11 days, so any given lunar calendar will generate an extra month about every three years, and there are lots of different lunar calendars.
(For example, the Chinese calendar occasionally repeats full months in order to keep the month of the year lined up with the season. Whenever this happens, there will be 13 months in the year, of which two share the same name.)
Even better, I'd prefer to have more semantic meaning, and for budget-2025-09.csv, buget-2025-10.csv to work everywhere...
Are filenames likely to include those representations? I feel like probably not (can you even include commas in Windows filenames?)
More to the point of the article--if you want things sorted by date, sort by date. I think most laypeople aren't looking at long CHAR1234_5678 filenames anyway, they're looking at thumbnails and dates.
The most common date format used in Europe uses period separators so can often appear in filenames. Commas are probably more rare. Things like versions are often fractional like v1.3 or v1.11 and can appear embedded in filenames.
> can you even include commas in Windows filenames?
Yes.
> Use any character in the current code page for a name, including Unicode characters and characters in the extended character set (128–255), except for the following: The following reserved characters:
< (less than)
> (greater than)
: (colon)
" (double quote)
/ (forward slash)
\ (backslash)
| (vertical bar or pipe)
? (question mark)
* (asterisk)
https://learn.microsoft.com/en-us/windows/win32/fileio/namin...
The entire idea that numbers would be treated on a character by character basis rather than as numbers is somewhat intuitive for developers and not for non-technical people.
The answer to all of those questions is no for lexicographic ordering. Lexicographic ordering leads to surprise and confusion as a result.
> It's like a feature request that initially sounds reasonable and useful but once you explore the requirements in detail you realize there are too many edge cases to be able to meet the request in a non-brittle way.
It's been on windows and macOS for coming up on 25 years, and is in practically every modern UI. It’s reasonable.
Agreed.What's more, the idea that people learn to put leading zeros is wrong and impractical, unless you know in advance how many digits you need. When you go from version 5.9.17 to 5.10.0 you don't go back and relabel every existing folder as 5.09.17.
The today standard way of sorting is well defined, unambiguous, and natural. Lexographic has its place, but user facing interfaces ain't it.
Had this in the Beat Saber mod manager recently. The game released 1.40.10 and my mod manager suddenly thought that game went backwards from 1.40.9
No
Just no
User interfaces that try to be cleaver are a pita.
Keep it simple, and avoid the confusion with corner cases that otherwise will baffle users. Like this
Lexicographic order is great when you need an unambiguous criterion that will work the same in every implementation; but you only need that for automated processing, i.e. for coding.
For user-facing presentation, having 5.9.xxx before 5.10.xxx is simpler; the corner case that baffles users is having 5.1 and 5.10 before 5.2.
LOL I can tell you don't have the experience of designing UI and shipping product to end users
> Keep it simple
What's simple? Good defaults make things simple, which means putting 9 before 10 in case, for the reason explained by parent.
> I agree with Microsoft/Google/KDE's order. The author's situation is extremely rare...
Even if that were a valid reason for making it the default behavior, the real issue is they don't even give you the option to have the lexically correct sort order. They just decided to give you something that's not accurate and that's all you get.
A trend which is frustratingly, increasingly common.
It's trivial to allow customization behind menus. But we rarely get that anymore. Especially for sandboxes devices like phones.
It's a giant middle finger to users who want to actually use their devices as a tool, instead of simply a portal for more sales and marketing.
I think the only problem is that it's a surprise and mystery, particularly because "dumb" alphabetical sort has existed forever. When they "fixed this" for the 99% of regular users cases, they should have made it as separate "smart natural sort" option separate from the "strict alphabetical sort" option (next to date, size, etc). Simple and obvious, rather than surprisingly different from the decades of experience that even non-technical users already have.
It's not just the one decision though; there are literally thousands, maybe tens of thousands, of these decisions in most software. You want every single one of them to have an option? You want it to support every single combination? At some point, it is ridiculous. Sometimes you just have to decide how your software is going to work and not leave every single decision to the user.
You don’t let every decision to the user, you make good defaults, but leave the option to override to the user! And thousands isn’t scary as long as groups/tags/search work, so what’s ridiculous about empowering the user?
Increasing the number of different possible combinations of settings your software can be running with by a factor of one nonillion is not a choice I’d make if I wanted to have any confidence in its reliability and security.
How the files sort seems kinda important. It gets at the core behavior of the program. It's not something superficial like a default icon, which the user probably can change.
It may be one of thousands of decisions, but it's one of a handful that are exposed in the user interface as a fundamental action.
In a file manager? Any more than the displayed thumbnails, icon size, whether folders are separated from files, whether images are separated from videos, what video types are supported, what file types are opened inline, what the click and double click behaviours are, etc?
And yeah kde has settings for all these but kde is also known for being too configurable.
I might be wrong on this, but I vaguely recall that on macOS back when you could commonly option-click to reveal advanced options, if you held option when clicking a sort it would change how it sorted from alphabetical to lexical or vice versa. I’m not a thousand percent sure of it, though, I think when I needed it I was able to set a directory preference via terminal to change how a specific directory was sorted and it was an option there. MacOS had (or has) a lot of buried options which I presume date back to its origins as a Unix as well as a convenience to its developers. A lot of the command line utilities were hacked calls to graphical settings code though, so it wasn’t very stable version to version as the UI calls changed and nobody prioritized non-UI bug fixes or breaking changes. These days CLI is nearly forgotten or assumed to be an exploit vector - see Screen Time data for example.
But the alternative would be a surprise to people who assume "by name" will order numbers, including those who are new to technology (and I think most non-technical people who sort things manually unknowingly order numbers).
We want to minimize surprises and mysteries, but computers have so much hidden complexity it's impossible to eliminate them. If users were shown a full description of how every feature on their computer worked before using it, they'd quickly start ignoring the descriptions. There should probably be a tooltip or "manual entry" for "by name" for those who are curious, and it should never be labeled "alphabetical" because it's not. But cases like the author's, where he assumes a feature works differently than most people (including the designers) assume, can't be helped.
> and the situation where someone wants "10" to be before "9" is far more common.
I guess you mean "after"? Otherwise it seems to me you're agreeing with OP.
> desktops don't label this sorting "alphabetical" (E: and it would really be "lexicographic"*), they label it "by name" (an informal criteria), so technically they're not lying.
FYI the more formal name for the "by name" order is "natural sort order".
> I guess you mean "after"? Otherwise it seems to me you're agreeing with OP.
Depends on which direction you're sorting in, no?
> Depends on which direction you're sorting in, no?
In a vacuum: yes. In this particular case: no, because we have the article's context clarifying that we're talking about ascending order.
It’s more confusing. I thought the article was correct when they said -10 coming before -9. Why? Because they were talking about the strict alphabetical sort. They are already prepending zeroes to force the comparison to be 10 vs 09. So, yes, they were talking about ascending order, but not natural ascending order, but ascii sorting order where 10 is before 9 because the comparison isn’t 9 vs 10, but 1 vs 9.
It was only clear to me because I could guess where they were going. They were complaining about natural sort vs alphabetical sort, which is a case I’ve run into many times, so I could see the argument coming.
The irony to me was that they were already altering how they named files to fit what they thought the computer wanted by prepending a zero to get a proper alphabetic sort. And even after that, some computers didn’t follow their idea of what it should be doing.
"The author's situation is extremely rare"
People sorting their files for alphabetical order is extremely rare?
And right now I fail to see even one 'case where someone wants "10" to be before "9"'
People sorting their files in alphabetical order but who want numerical values in their files to be sorted digit by digit instead of as numbers is the rare case.
I might go further in my ideal sorting algo which would be normalize capitalization and ignore all alphanumeric characters and treat them all as separators.
You mean file9 before file10?
I have some beef with microsoft, that you can only change this at the Computer level, not per user (see registry key below). Also they call it natural sorting for users, but logical sorting internaly. Unify your termini!
[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\CurrentVersion\Policies\Explorer] "NoStrCmpLogical"=dword:00000001
To change it per user, set it in the user's hive instead of in the local machine hive (e.g. HKEY_CURRENT_USER instead of HKEY_LOCAL_MACHINE)
TIL they are called "hives". Windows Registry is an interesting thing. Even casual users have to interactive with it once or twice w/o fully understand it.
https://learn.microsoft.com/en-us/windows/win32/sysinfo/regi...
Raymond Chen explained why a registry file is called a “hive”:
Because one of the original developers of Windows NT hated bees. So the developer who was responsible for the registry snuck in as many bee references as he could. A registry file is called a “hive”, and registry data are stored in “cells”, which is what honeycombs are made of.
https://devblogs.microsoft.com/oldnewthing/20030808-00/?p=42...
https://devblogs.microsoft.com/oldnewthing/20030808-00/?p=42...
Thanks for fixing the link!
The file sorting isn’t something relegated to niche users because of the prevalence of tv episode file name sorting (eg S01e01) and it has necessitated the leading zeroes to make it work properly with “alphabetical sorting”.
And that would sort correctly with both methods, though, especially when each "field" is delineated (e.g. Show.S0XE0Y.Episode.Name.HEVC.1080p.mkv)
Both should be supported.
Perhaps put the uncommon (true alphanumerical order) behind a nested menu or something. But the mind-reading-less option should be there.
> Both should be supported.
At least in KDE they are, and you can pick whether you want natural or alphabetical sorting (which has a case sensitive and insensitive variant).
What is the reason to append a textual file name with a number? User Experience?
They are magic numbers. Maybe a serial ID, date stamp with more magic, revision, release, ...
Magic Number land has 10 > 9 in the above.
9 > 10 is only possible when removing the Magic Number and morph into mealiness text.
At the moment I cannot think of any magic number where 9 > 10.
>You may be looking at that time through rose-tinted glasses.
Nope, regarding what he talks about, the time was rose-tinted itself.
This is reminding me of the whole "Worse is better" essay and debate:
https://news.ycombinator.com/item?id=27916370
The author wants the "worse" sort, one based on ASCII/Unicode codepoints, without any intelligence for numbers that 99% of GUI users want.
For their purposes, they've assumed something about the implementation, to the point that a convenience feature is actually a misfeature for them. But the author here is probably a developer, or close to one, so they do not represent the needs of most people using computers.
Understanding the target audience for your product results in very different design decisions. Better is better might be great for products, but worse is better is probably better for systems that need to grow and evolve.
It's an issue of mental models. As a developer, his mental model is one of how naive software would sort items with mixed numbers in them. Most people, of course, naturally sort 10 after 9 -- their mental model doesn't contain software developer assumptions.
> The author wants the "worse" sort, one based on ASCII/Unicode codepoints, without any intelligence for numbers that 99% of GUI users want.
I want the author's opinion on how caplital and lowercase letters should be sorted. Do they follow strict ASCII/Unicode codepoints, or do they normalize into actual alphabetical order and sort upper/lower within each letter?
> I want the author's opinion on how caplital and lowercase letters should be sorted. Do they follow strict ASCII/Unicode codepoints, or do they normalize into actual alphabetical order and sort upper/lower within each letter?
I prefer the strict ASCII / Unicode sorting (all capitals first, then all lowercase).
And where do you sort the letter ä? (After a is correct in German, but I think Swedish does it differently.)
This feels like the right moment to mention "ch", which is considered a letter in orthodox Czech, sorted between "h" and "i". The problem is, you can't reliably distinguish between "ch"-the-letter and "ch" as just "c" and "h" combined, which are present in loan words but also some original Czech compound words.
So if you're doing it "properly", sorting strings in Czech involves understanding the etymology of every word.
That's why we have all this LC_* stuff in Linux, which you can configure to your needs:
Mix in your Swedish or Swaheli, maybe even the Vatican State:Asciibetical sorting
>Understanding the target audience for your product results in very different design decisions
This is an excuse. Just add an option to sort both ways. It isn't hard.
There is no target audience in this planet that benefits from less options or less features. Even if you had the features under an "advanced mode" UI that's still a better software than not having the feature in first place.
Have people forgotten the 80/20 rule? Most features will be used by only a small slice of users, that doesn't mean they're out of scope.
Sorry, I'm just kind of exhausted of software not being able to do the most obvious things because it didn't align to some perfect vision of how the user should be.
The most irritating circumstance for this is looking for files named with a hash:
This is one of the settings I immediately turn off on Windows via the registry key mentioned in the other comments here.I miss the time when computers did what you told them to, instead of trying to read your mind.
These days, it's more like "trying to change your mind". I absolutely hate the "the user is wrong" authoritarian mentality that unfortunately has infected a ton of software, even open-source.
Exactly. This is even more annoying when it isn't exactly a hash, but some gibberish you cannot really make sense of, which does have a numeric section in them: like a user ID, or unix time, or who knows what else it could be, but you are trying to visually find a file abcd89764237 somewhere after abcd683426834, and it isn't evident why you cannot, unit you notice that the latter has more digits in its "ID" for some reason.
I almost always want the version-sorting that's being presented in this article, rather than an "alphabetical" sort. But on the other hand, it absolutely seems like a valid bug that this is presented as an "alphabetical" sort, rather than something like "alphabetic/numeric" or similar. In other words, a problem of labeling rather than one of sorting.
It’s not being presented as an alphabetical sort, though. The author assumed that sorting by name meant an alphabetical sort, but that’s not how it’s labeled.
In fairness, sorting by name has, for many years, been an alphabetic sort. Doing a mixed alpha/numeric sort is a relatively new thing.
Yeah, exactly. The behavior described is actually very useful. The problem is imposing it on the user with no warning or option to turn it off.
Author here - I Agree both with you and with the parent's comment. Having two options in the "sort by menu" - like "Name (natural)" and "Name (strict)" or something - would have solved everything.
> The problem is imposing it on the user with no warning or option to turn it off.
You can say that about every single design decision made about every product.
The gripe about this particular feature seems misplaced because almost all users will want the sort that's offered and the actual alphabetical sort is likely the desire of a more advanced user who, in fact, is offered a choice through registry editing and/or using a more advanced cli option for the occasion they might need an alternative sort.
This is a sensible default.
> You can say that about every single design decision made about every product.
No, that's not true. Many aspects of my computer's UI are user-configurable.
Yes but not every single one of them
Obviously. All I'm saying is that this particular decision ought not to have been taken from the user. Real alphabetical order is not an unreasonable thing to want.
“Real alphabetical ordering” is incredibly nonspecific. It’s underspecified even for ASCII-US, but essentially meaningless for those of us in 2025 who need to handle Unicode.
How do capital letters sort relative to lowercase letters? How do letters sort relative to digits? How do you consider code points that can correspond to different letters in different lettering systems with different ordering? How do you handle diacritics? Do you want the behaviour to be stable through Unicode normalization? Should it differ based on the character encoding? Should different representations of the same character, such as blackboard lettering or circled numbers, be sorted with other representations of the same character or grouped separately?
You can come up answers for these questions, but there’s no unambiguously correct option. The least subjective option is sorting based on encoded byte representation (if that is even specified), but that is not “alphabetical” and would not be intuitive to most users.
The issue at hand is how numbers are sorted. That has nothing to do with unicode.
Unicode has many different representations of digits, and I would dispute using the term “alphabetical” to refer to digit ordering in any case.
Notably, some versions of “sort” on Linux have version sort nowadays. sort -V
I actually don’t know exactly how it works internally and it is a little bit magical, but I use it all the time when looking through my files because it just sorta works in most cases. Of course a nice thing about it is easy to turn on or off.
The term for the sort in the article is called lexical, but the problem is the people are stupid.
The average user does not know the difference between lexical and alphabetic sort
https://www.unicode.org/reports/tr10/#Contextual_Sensitivity:
“There are additional complications in certain languages, where the comparison is context sensitive and depends on more than just single characters compared directly against one another,
[…]
Numbers. A customization may be desired to allow sorting numbers in numeric order. If strings including numbers are merely sorted alphabetically, the string “A-10” comes before the string “A-2”, which is often not desired. This behavior can be customized, but it is complicated by ambiguities in recognizing numbers within strings (because they may be formatted according to different language conventions). Once each number is recognized, it can be preprocessed to convert it into a format that allows for correct numeric sorting, such as a textual version of the IEEE numeric format.”*
I think those file browsers made the right choice, even given that they don’t (as in this example) always do the right thing.
But -10 is smaller than -2, right?
Filenames rarely have negative numbers in them, and it'd usually be ambiguous whether they were negative or dash-separated positive.
I thought this was pretty well known. E.g. the macOS Foundation library even exposes NSString.localizedStandardCompare() [1] which implements the sorting algorithm used by Finder, and should be used by any well-behaved macOS application. Windows uses StrCompareLogical [2].
[1] https://developer.apple.com/documentation/foundation/nsstrin...:)
[2] https://learn.microsoft.com/en-us/windows/win32/api/shlwapi/...
I would have assumed it worked the same as ls, so I found the article interesting. But now that I know, I think this way is better.
I can’t think of any case where I would need purely alphabetical sort. In most photo browsing apps, photos will be sorted by timestamp rather than filename. If I really needed it to sort properly in file explorer, I would try sorting on created date. And failing that I would probably just normalize the file names.
> Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means. So when you ask them to sort your files alphabetically, they don’t. Instead, they decide that if some piece of the file name is a number, the real numerical value must be used.
Well, no. You don't actually ask them to sort in alphabetical order. You ask them to sort "by name", and that is up to their interpretation. And they choose the interpretation that (per their reasoning, and possibly some actual data) seems most likely to correspond to what the user wants.
Maybe future versions of those OSes will add a rule that says that if any of the number groups have leading zeros then it reverts back to actual alphabetic order. Or maybe they'll give you configurable options. (Maybe some of them already do.)
Clearly a leading zero means the number is in octal (but only if all the subsequent digits are between 0 and 7). I think that would lead to the most intuitive results.
> And they choose the interpretation that (per their reasoning, and possibly some actual data) seems most likely to correspond to what the user wants.
Yes, that make sense, but the problem is that this interpretation changed in the last 10 (15? 20?) years. It used to be that "by name" meant "by name, il alphabetical / lexicographical order" in pretty much every file manager.
Reminds me of https://xkcd.com/1172/
Sorting so "foo9" is before "foo10" is called natural sort. I found out about natural sort a week ago and I am thrilled that my programs now print their output in a sensible order. Give natural sort a try and see if it improves your life too :-)
I found the magic two lines of Python to do a natural sort here, by the way: https://stackoverflow.com/questions/11150239/natural-sorting...
Natural sort is an Option in sort(1):
And we have "sort -h" to sort the output of e.g. "du -sh *" properly.Edit: formatting and add sort -h
"I created the Alphanum Algorithm to solve this problem. The Alphanum Algorithm sorts strings containing a mix of letters and numbers. Given strings of mixed characters and numbers, it sorts the numbers in value order, while sorting the non-numbers in ASCII order. The end result is a natural sorting order."
https://web.archive.org/web/20210207124255/http://www.daveko...
There are many older instances of that, such as "versionsort" from various Linux tools and libraries. I think this has likely been independently recreated several times, with various subtle differences.
I get it, but if all these major operating systems are handling this same ambiguous [0] situation in the same way, perhaps one needs to reevaluate their mental model or expectations.
Am I out of touch? No, it's the operating systems who are wrong
0 - numbers are not part of the alphabet.
ls sorts filenames strictly lexicographically, comparing character by character, so e.g. "055436307" is compared as the characters "0", "5", "5", etc. so it sorts before "121134" because "0" is less than "1". if all compared characters match and one string ends, the shorter one comes first. Symbols like _ are just more characters, and their position relative to digits depends on the locale’s collation table.
Google Drive uses ICU collation with the numeric option enabled, which treats each consecutive sequence of digits inside the filename as an integer. so "055436307" is parsed as the number 55,436,307, while "121134" is parsed as 121,134. and since 121,134 < 55,436,307 then "121134..." comes before "055436307..." even though lexicographic order would suggest the opposite. and i think when two digit runs have the same numeric value, the shorter run comes first; if runs are equal and the string continues, then normal character comparison resumes, including any underscores or suffixes
It feels like this algorithm could be improved though. If a number has leading zeros you probably don't want to sort it numerically.
That said the author's situation where it's numerical and different lengths seems likely rare enough that it probably isn't worth complicating things.
Earlier this year I submitted a bug to VSCode about sorting Playwright tests in the alphabetical-and-numerical order that VSCode favours, after Playwright told me it was a VSCode issue.
Some people rushed to fix this as I'd done some diving into the issue and presented the relevant information and code, so now VSCode's Playwright test list uses the same sorting mechanism as the rest of VSCode.
Sadly, the underlying Playwright does not receive that order from VSCode so it still actually runs sequentially-numbered tests in strict alphabetical order. :(
Well, lots of interfaces don’t say “alphabetical” anymore, they say “name” or some variant, and then they can define it however they want, regardless or because of the frustration it causes users but not some other users which will now be inverted for long term-frustration averaged user experience.
Maybe it's just me but I don't miss this at all:
The only time natural sort bit me was with nonsensical names like <md5>.jpgNah that's not just you. That is an unnatural way to sort things because that's not how numbers are ordered. I remember when Windows changed to sorting numbers by their value and, despite my programmer brain finding it strange in a way, I was super happy to have files display in an order that actually made sense.
I think it depends on the person. That order is exactly what I expect and want.
Isnt the author confusing "alphabetical sorting" with "ASCII sorting"?
Afaik there is no universal way to handle numbers in alphabetical lists. Sometimes numbers some before letters, sometimes after, etc.
A digit is not a part of the alphabet, right?
Numbers aren't in the alphabet. So no, you don't mean alphabetical order.
I felt a little bad about this snark but actually, author barely understands their own use case (says they want alphabetical order but they actually want something more) and barely understands the UI they're using (says they asked for alphabetical order but none of the file managers they used says it has any such setting) and then they go on to claim this is to satisfy dumb users:
> Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means.
The so-called "natural" sort makes sense for version numbers and enumeration (without zero-padding) but I'm more often dealing with file names with a datetime (like in the article), a hexadecimal hash, or just randomized string of characters that includes numbers. In those cases "natural" sort makes it harder to find the file you're looking for.
Even when files are enumerated it's pretty rare to have more than 9 parts and no zero-padding, whereas there are almost always multiple consecutive digits in the use cases for which "natural" sort is not a good fit for. It just feels like a bad default, at least for a programmer's workload.
Dumb tools are more robust.
To answer the question in the article, I’m pretty sure Windows Explorer (and probably File Manager before that) has sorted filenames this way for at least 30 years.
I can confirm that this does not happen in Windows 98, but does happen in Windows XP.
"The Tyranny of the Marginal User" strikes again: https://nothinghuman.substack.com/p/the-tyranny-of-the-margi...
I got used to naming files/folders with leading zeros when I want them to be sorted alphabetically (for example payslips/invoices, etc).
But I'm a tech guy, I know what does "alphabetically" mean in the tech world. And it probably is not what common folks mean when they think "alphabetically" outside the tech world.
Edit: in fact, if I recall correctly, the proper term for this kind of sort (the one OP wants) is alphanumeric sort.
I also got used to it, but especially when writing short scripts that generate numbered files it gets annoying to have to pad with zeroes every time, and also precommit to a specific amount of digits you want to allow (finding a compromise between adding a ridiculous amount, like 20, and using only 4 despite knowing the script might one day surpass 10⁵ files).
The natural numbers are ordered. Let me use its ordering instead of having to rely on an ad-hoc lexicographic fixed-length tuple representation of decimal digits, without any padding. My position is that numbers in filenames should always be considered atomically unless explicitly instructed otherwise.
If there were no issues of backwards compatibility, I would thus advocate for changing ls. Eza (maintained fork of Exa, Rust-based ls alternative) actually does sort this way by default, much to my delight.
shameless plug (though I don't get a cent out of this, of course): https://blog.vslira.net/2025/03/a-neat-approach-for-sortable...
There's a Group Policy setting in Windows: Computer Configuration\Administrative Templates\Windows Components\File Explorer\Turn off numerical sorting in File Explorer Group Policy has so many essential settings I hurry to change with every isntall. I wish Windows would expose more of them to the user in ordinary settings.
I think if we (in our industry in general) had REAL agile, and not pseudo-waterfall "the designers design it, the engineers implement it, QA QA's it, and then we lay everyone off because they're no longer needed" (but, loophole alert! We did daily standups and used Jira, so it was "agile" the whole time!), then we'd have a snowball's chance in hell of actually having a reasonable solution to this. Off the top of my head, this seems like something that should be a setting in control panel. But, because everyone assumes (contra to agile) that the designers "got it right the first time", this kind of improvement can't happen.
That’s why I don’t even bother with the file name for photos.
1. Sync all equipments to the same clock.
2. Sort by Date Taken, if unavailable, sort by Date Created.
Yes, sounds to me like the user really wanted to sort by time created. And got used to sorting alphabetically as a poor proxy for that.
When copying files from a device and then between systems, too often the dates get lost (shouldn't, but still...)
That only happens for the datetime metadata of the files (modified, created, access etc). The EXIF metadata will still remain the same.
Plex team, are you reading this?
For some inexplicable reason, Plex just throws its hand up on non-ASCII characters and puts them first.
In Norway we have three extra letters, æøå, and they're at the end of the alphabet after z. But in Plex, I have Øystein Sunde[1] placed before any other in my music library.
Now in the 1990s I would forgive US software for such a thing, but it's 2025...
[1]: https://en.wikipedia.org/wiki/%C3%98ystein_Sunde
If only we could represent sort order by some structural form of decision logic which also embeds encoding a regular .. well.. expression matching a pattern..
I think the real issue here is that two Android phones take photos with incompatible naming schemes.
I am sure that at some point someone thought the milliseconds should or should not be separated from the seconds and made that change without thinking through the consequences.
The correct sorting algorithm is described here: https://manpages.debian.org/stretch/dpkg-dev/deb-version.5.e...
I was joking. Really I would sort file names lexicographically. But the way Debian sorts version numbers is interesting and seems like a good way of handling that particular situation.
Sort by the time the photo was a taken in the metadata?
Why is the author so perplexed by it?
https://en.wikipedia.org/wiki/Natural_sort_order
It's simply natural sorting. I don't see what is so controversial about it.
I find the term "natural" to be inadequate here, there is nothing natural about sorting strings in this particular fashion compared to another. It should be given a more descriptive name, like "Number-aware alphabetic" or something like that as to actually give a hint about what it does.
Have you tried sorting by created date instead? What is the point of relying on filename if that's not what you want.
Heh. One of the bugs that once caused me to bang my head against the wall was caused by the Estonian language. Its alphabet has Z following S and Š. So the "foolproof" regexp to match the letters '[a-za-Z]' was misfiring for some entries.
Be glad you don't have to deal with non-ASCII characters: acute/grave/tilde/umlaut/diaresis/etc. accented characters, dotted vs. dotless 'i' (Turkish), barred i (an 'i' with a sort of dash through the middle, used in some languages for a sort of schwa-like vowel), thorn, not to mention non-Roman characters. And different languages sort the same characters differently, so you can't just pay attention to their Unicode values. (@cubefox has a post here pointing to the Unicode Consortium's doc about sorting)
When I say Afferbeck Lauder [1] I mean "alphabetical order"
[1] https://en.m.wikipedia.org/wiki/Afferbeck_Lauder
Those are beautiful, thank you for posting them
Renaming things to make them queue correctly (I usually couldn't care less about visual sorting, I use a terminal) is by far my #1 task by LoC and frequency of occurence, and by far the most annoying. Metadata can be very helpful to obviate this issue, but it usually just leads to another problem where you now need metadata editors and readers in addition to the "user-visible" name metadata. It's frustrating.
I think the algorithm is probably incorrect. A number starting with 0 should be treated lexically not numerically. Otherwise you have a situation where img_1_01.jpg and img_01_1.jpg does not have a complete ordering.
That's not the issue.
The issue here is that one camera appends milliseconds to the seconds without a separator, and the other uses a separator.
So of course the ones that include milliseconds look like bigger numbers and get sorted last.
Leading zeros aren't the issue here.
> Otherwise you have a situation where img_1_01.jpg and img_01_1.jpg does not have a complete ordering.
(Good) "natural sort" implementations generally have ways of handling ties like this. It's similar to the problem of case-insensitive sort over case sensitive sets.
It wouldn't be the first time widely-used software sorted numbers by a function that does not produce a total ordering. For example, Excel: https://gregat.es/excel-numeric-order-transitivity/
> But nope, this is not it, because the good old ls sorts my files correctly
Did the author try "ls -v"? It would probably give the exact same order these file managers used.
I'm often yelling at the software on my computer: "STOP TRYING TO HELP ME!"
It's like having a toddler help you make a meal. It wants to be involved and recognized so badly. Meanwhile I'm starving and just want to get the food done as quickly as is possible and I'm constantly tripping over this little ball of misguided efforts.
Please. Stop trying to be smarter than me. You often can't, and when you get it wrong, you make it measurably worse. If you insist on doing this please give me the "Expert Mode" setting back so I can flatly disable ALL OF IT with one click.
If you don't like the default natural sorting order, you can just change it in Dolphin. Settings > Configure Dolphin > View > Content Display > select anything other than "natural". You can even pick if you want case sensitivity or not.
The OS doesn't think you're too stupid to understand sorting, it relies on you being smart enough to figure out where the setting is located. In this case, four levels deep is probably too much to ask from users if they will write an entire blog post like this before finding the toggle.
I thought this was going to be a deep dive into what "alphabetical" means and how that's itself not a universal term between locales, what with so many different collation preferences.
That would likely have been a more useful article for the average developer. It is extremely hard to be aware of all the ways strings of different locales can defy our intuitions.
This got me riled up to the point where I blew a gasket and just can't..
I agree with the article.
I liked computers better when everyone hated them, and for the reasons they hated them..
I agree that the base functionality of just sorting character by character can be occasionally useful. However I would really be interested in seeing why you believe this to be the correct choice for user-facing graphical file managers, as its evident problems with typical usage seem more salient compared to the edge cases as illustrated in the article.
When you get a bunch of files (let's say 1000+) without leading zeroes, this is a blessing. But I get the author's frustration, the expected behavior is not there, instead, he gets magical sorting that is wrong for his use case. I'm not sure what the ux should be, and maybe the algorithm here could be smarter, but it's a trade-off.
> the expected behavior is not there
The expected behaviour is ambiguous (and thus subjective). Older versions of windows shipped with alpha sort. New versions ship with natural alpha sort. According to the UX designers over at Microsoft (and surely the user feedback), natural sort _is_ the expected behaviour.
I certainly agree with natural sort being the expected behaviour too.
>Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means.
i really really hate this framing, and i see it far too often. no, the operating system developers did not make a value judgement about their users. they observed their users to find out what behaviour was expected, and they designed the behaviour of the system to match the behaviour that the majority of users expect.
and then you made an incorrect assumption about how the system works, and decided that your incorrect assumption means everybody else is dumb and you're the only smart person in this situation?
Even if you are a file naming Einstein and you always zero pad your integers to exactly right length, we have this thing called the internet, where you can download other people's files.
Your OCD is not my OCD.
This is one of the big ways that LLMs are going to change the game for UX. Your operating system is going to have some sort of 'butler', which knows all of your preferences, and the butler will go through the APIs and man files and informational dialogs of every app you use and auto-configure them.
Then if you want something to change, just ask the butler. If the app is open source and doesn't support the requested feature, the butler might even be able to code it up.
Bro forgot that ls has an option to obtain the same sorting as the one he doesn't like.
ls has it because it solves a real user need.
This was a fun thing to realize in my early days of programming in Delphi. I guess the author will soon realize why old systems name things ticket "00001" and so on.
I've encountered a tangential problem to this with package versioning on Linux distros. Thankfully it was not too hard to write an algorithm to compare versions (thanks AI!).
Another problem which annoys me to no end is that most file managers and file selection boxes put directories before files.
This makes it hard to find the file that was most recently changed, for example. Which is an action that is extremely common. (In fact, why does my file manager not have a most-recently-used shortcut?)
I rename all of my photos upon import using the created date, formatted as `YYYY-MM-DD kk:mm:ss`.
But it would frankly be great if most file browsers just let me sort photos based on metadata. But then I just end up in a dedicated photo browser, instead.
Por qué no los dos?
Call lexicographic order "sort by name" as it's called now, and call dumb character-by-character sort "plain" or something like that. I'm not a designer, maybe there are more intuitive names, but come on. This isn't an intractable problem.
I have the same problem on Nemo. More specifically, I had made a small app that displayed files of a directory in alphabetical order, and then when I look at it in Nemo it isn't the same order because I didn't implement their smart algorithm.
I fail to see how this is a "problem"? You implemented a sorting mechanism that was useful to your application, while Nemo implemented another which as this thread demonstrates seems to be much more useful and intuitive for the average user. This is also of course not specific to Nemo, as no 'modern' file manager on Linux sorts filenames like it's 1980 and all you are able to feasibly do is step through the bytes.
Nemo isn't some random app a teenager made. It's the default file manager of a desktop OS. I expect it to cover more use cases.
I expect the same for other file managers on Linux. Although I must say I'm generally let down by Linux software.
> But 1 is smaller than 9, so file-10.txt should be first in alphabetical order. Everyone understands that, and soon people learn to put enough leading zeros if they want their files to stay sorted the way they like.
No. Not “everyone understands that”. Natural sort happens in real life and everyone understands that. Only those who understand ASCII — not the average user of graphical file managers — will deduce the reason for your definition of “alphabetical order”.
> Now that I know what the issue is, I can solve it by renaming the files with a consistent scheme.
Intensely ironic given the previous suggestion.
In Total Commander, there is a function in the options to sort strict by numerical char code. It will sort those files correctly. Unfortunately, it will also sort "10.txt" before "2.txt".
---
In all file managers, I miss an API point where one can give a userdefined sorting function for the file and folder list.
What do you mean by "Unfortunately"? This appears to be the only correct conclusion from the algorithm you selected, you can't eat the cake and have it too.
Regarding your second point, that's not really what a graphical file manager is for, I think. At this point (likely even earlier) you would be better off just writing a simple script in the scripting language of your choice. (If going for something fancy, you could also implement a FUSE based on symlinks for the original files, where the filename is prepended by a sort key. This would work for every major file manager and you could manipulate the files in mostly the same way as before.)
Paragraph 1: I speak of a sorting method which splits the filename at the boundaries between numbers and non-numbers, and sorts by the parts of the resulting tuple, the numbers naturally (10 comes after 2) and the rest by numerical char code.
Paragraph 2: I am not sure what you mean here with writing a script. The graphical file manager shall sort its file list using the sorting function I hand over to it.
"That's not really what a graphical file manager is for". Says who? Every software which has a plugin system does that, why should a file manager not?
Ha! I had the exact same realization on MacOS. Extremely annoying behavior.
Sorting by name (collation) is waaay tricker than simply figuring out how to parse the numbers.
The International Components for Unicode library implements the Unicode Collation Algorithm, which depends on the language code and region of the locale, and looks up the quirks for each locale in the Common Locale Data Repository.
It's a much better idea to just use the standard ICU library or platform specific libraries (which are often build on ICU like JavaScript's Intl.Collator), instead of trying to hot dog it by rolling your own.
International Components for Unicode
https://en.wikipedia.org/wiki/International_Components_for_U...
>ICU provides the following services: Unicode text handling, full character properties, and character set conversions; Unicode regular expressions; full Unicode sets; character, word, and line boundaries; language-sensitive collation and searching; normalization, upper and lowercase conversion, and script transliterations; comprehensive locale data and resource bundle architecture via the Common Locale Data Repository (CLDR); multiple calendars and time zones; and rule-based formatting and parsing of dates, times, numbers, currencies, and messages.
Unicode Collation Algorithm
https://en.wikipedia.org/wiki/Unicode_collation_algorithm
>The Unicode collation algorithm (UCA) is an algorithm defined in Unicode Technical Report #10, which is a customizable method to produce binary keys from strings representing text in any writing system and language that can be represented with Unicode. These keys can then be efficiently compared byte by byte in order to collate or sort them according to the rules of the language, with options for ignoring case, accents, etc.[1]
>Unicode Technical Report #10 also specifies the Default Unicode Collation Element Table (DUCET). This data file specifies a default collation ordering. The DUCET is customizable for different languages,[1][2] and some such customizations can be found in the Unicode Common Locale Data Repository (CLDR).[3]
Common Locale Data Repository
https://en.wikipedia.org/wiki/Common_Locale_Data_Repository
>The Common Locale Data Repository (CLDR) is a project of the Unicode Consortium to provide locale data in XML format for use in computer applications. CLDR contains locale-specific information that an operating system will typically provide to applications. CLDR is written in the Locale Data Markup Language (LDML).
>Among the types of data that CLDR includes are the following:
Tricky collation examples:sv-SE (Swedish): å, ä, ö are separate letters at the end of the alphabet, not variants of a or o.
de-DE (German): ä, ö, ü may sort as ae, oe, ue in some contexts, or as distinct letters. ß sometimes sorts as ss.
tr-TR (Turkish): dotted i (i) and dotless ı are different letters; I sorts with ı, not with i.
es-ES (Spanish): traditionally ch and ll were treated as single letters with their own place in the alphabet.
cs-CZ (Czech): ch still counts as a unique letter, sorted after h.
da-DK / no-NO (Danish/Norwegian): ø comes after z.
is-IS (Icelandic): þ (“thorn”) is part of the alphabet, after z.
fr-FR (French): accents usually ignored in sorting, so é = e, but not always depending on collation settings.
el-GR (Modern Greek): tonos accents, final sigma ς vs. σ, etc.
nl-NL (Dutch): the digraph “ij” is often treated as a single letter, and capitalized as “IJ”. In dictionaries and phone books it often sorts as a single letter under “I”, but sometimes is listed after “X” depending on tradition.
Then you get into non-Latin languages like, Chinese, Japanese, and Korean collation, which gets hairy with radicals, kana order, and stroke count.
Also different locales have different ways of representing numbers, like switching between "," and "." as separators and decimal points.
ICU supports integer only "natural" numeric collation, so anything more complicated like versions, floating point, negative numbers, hex, thousands separators, fractions, roman numerals, etc, you'd have to build on top of ICU.
ICU doesn't support incomprehensible dead languages like Latin or Ancient Greek (it does however support French ;). It does support Roman numeral formatting, but not collation, which would be pretty tricky and ambiguous.
https://www.youtube.com/watch?v=sKWvTlLMB-Y
A nuanced but common example that ICU/UCA/CLDR helps with is a menu to select the current locale: you have to translate each language's name into the current locale, and also sort them in the current locale. On top of different collations they can also have totally different spellings, like "United States of America" is "Verenigde Staten van Amerika" in Dutch. This makes it challenging for users to find their own language when the locale is set wrong! You just can't win.
Not to mention emojis! Which comes first: The chicken or the egg? The taco or the poop?
Also, the Mac Finder switches ":" and "/" for historical reasons (HFS used to use ":" as a directory separator instead of "/"), so you can create a file name like "9/11 Attack" in the Finder, which actually gets the underlying Unix filename "9:11 Attack". Don't believe me? Rename a file in the Finder to include a slash, which you know is impossible to represent as a Unix file name. Then go "ls" the directory in the shell.
The Mac Finder weirdly collates "/" after "9" because under the hood it’s really storing it as ":", which sorts before "0". But it also has other punctuation collating inconsistencies, sorting "," and ";" and others after "0" too. Definitely not ASCII order -- I'm not sure what rules it uses, but it's different than "ls".
However, while it's generally true you can't have "/" in Unix file names, NFS used to trustingly let clients rename Unix files to include a "/" in their name, which the Gator Box AppleTalk/Ethernet gateway let you do with the Mac Finder (pre OS/X), which would silently corrupt your "dump" backups on the Unix NFS server, so you would not learn about it until you tried to retrieve your files and "restore" crashed.
https://news.ycombinator.com/item?id=31821646
>Another reason that NFS sucks: Anyone remember the Gator Box? It enabled you to trick NFS into putting slashes into the names of files and directories, which seemed to work at the time, but came back to totally fuck you later when you tried to restore a dump of your file system.
>The NFS protocol itself didn't disallow slashes in file names, so the NFS server would accept them without question from any client, silently corrupting the file system without any warning. Thanks, NFS!
By the way, there seems to be a "standard" way to sort strings:
> Unicode Technical Report #10 also specifies the Default Unicode Collation Element Table (DUCET). This data file specifies a default collation ordering.
https://en.wikipedia.org/wiki/Unicode_collation_algorithm
I assume this mainly aims at giving a reasonable compromise between the different dictionary and phone book sorting rules of various languages (and even locales), which should give reasonable results for most languages. I assume this also puts "Alice2" before "Alice10".
> I assume this also puts "Alice2" before "Alice10".
It doesn't (per https://www.unicode.org/reports/tr10/#Non-Goals):
Oh, that surprises me.
> Of course, the user who named those files probably wants file-9.txt to come before file-10.txt. But 1 is smaller than 9, so file-10.txt should be first in alphabetical order. Everyone understands that, and soon people learn to put enough leading zeros if they want their files to stay sorted the way they like. Well, apparently all these operating systems have decided that no, users are too dumb and they cannot possibly understand what alphabetical order means. So when you ask them to sort your files alphabetically, they don’t. Instead, they decide that if some piece of the file name is a number, the real numerical value must be used.
I think there are many things wrong with your assessment of the situation.
First, where does it say in these file managers that they're sorting by alphabetical order? I see that you've specified that you want the files sorted by name, but I don't see that you've specified you want them sorted by name alphabetically. And what does "alphabetical sort" even mean when you're sorting characters which are not letters? What you mean is probably "lexicographical sort".
Second, you admit yourself that users probably want natural sort. Why would you expect these products to do the thing which they know users usually don't want by default? That just seems like bad design to me. They know users usually want natural sort, and you know users usually want natural sort, so why would you expect the default behaviour to be a lexicographical sort?
Third, just like how you've learned to work around the lack of natural sort in poorly designed products of years past by adding leading zeroes, you can just add trailing zeroes to get the lexicographical ordering that you want. Why do you seem to be implying that the latter is more user-hostile than the former? It doesn't make sense to me. A decision had to be made about what sort to use and they picked the one that most people want. Isn't that what we should be expecting in a product that caters to its users?
I see in other comments you've suggested that there should be a separate option for choosing between lexicographical sort and natural sort. But in the past, when lexicographical sort was the only option, why weren't you complaining about it being user-hostile to only have one option then? Why is it only when the default is something you're personally not used to that it warrants complaint? And where do we stop, do we have separate controls for every single sortable string field to determine whether it should be sorted lexicographically or naturally? Or just the name field? Don't you think that is going to lead to interface bloat?
TLDR: the author found out about natural ordering [0], i.e. treating a sequence of digits as a number while sorting.
Usually preferable, except when not. Just like distinguishing between upper- and lowercase letters, and other misery.
[0] https://en.wikipedia.org/wiki/Natural_sort_order
If I understand the article, the author wants magic :)
I take it to mean they want the system to know file_9.txt is less then file_10.txt.
I never saw that happen in any OS, so I do not know what he is referring to. Maybe whatever that old system was, it sorted by create time as opposed to file name.
So, the author can try and create "aisort" that will look at all file names and add leading zeros to the file numeric portion, sort, then remove the zeros added. That will probably as slow as s***t and use gobs pf memory, depending on the number of files.
No the author is saying the opposite. They expect file9.txt to be after file10.txt, but it many modern operating systems, it isn’t!
Really, I do not know how I missed that :) I read it a couple of times to and I still thought he wanted it the other way.
So my original comment kind of stands but in a opposite way.
I have never see file_9.txt sorted before file_10.txt, I just tested it on OpenBSD and I got this, which I have always seen:
$ ls|sort
file_1.txt
file_10.txt
file_12.txt
file_2.txt
file_20.txt
file_3.txt
file_9.txt
Author here - My surprise stems exactly from the fact that for the last few years I have exclusively managed my files via a the UNIX shell, which behaves in the classical way.
When I started using Linux as my daily driver after many years of Windows (but with familiarity with UNIX systems going way back), I knew it would be like that in the terminal, but it still took some adjustment. But actually, Nemo does the same "natural sort" thing, and also sorts case-insensitively.
>I take it to mean they want the system to know file_9.txt is less then file_10.txt.
The polar opposite, actually.
It’s not magic. It’s called natural sort and it doesn’t require gobs of memory. Most (all?) modern OS file managers will natural sort on file names.
That's not what the author says- they said that file managers actually are somehow sorting file-9.txt before file-10.txt, and it's breaking real alphabetical ordering.
i think it's the opposite, that they _want_ file_10.txt to come before file_9.txt by default, but that file explorers fail at this. it's rare that i want true alphabetical sort, but it's convenient for cases like tfa where alphabetical sort is more predictable if i have filenames that look like <letters>_<numbers-of-same-length>.txt.
Nope, you got it completely reversed.
Digits and any other characters that are not A-Z and a-z should not get sorted. That's the true result of doing what you asked and not what you meant. Pedantic, but that's why we are here.
No.
this is an ID-10T PEBKAC ERR.
Not this keyboard not this chair, but the problem is with idiots between keyboards and chairs.
The author is not the ID10T it’s the other general users.
The author is intelligent enough to recognize that this is not alphabetical sort, but the term that they are looking for to describe the sort that they see in dolphin windows, google etc. is *lexical* sort, not alphabetical.
The engineering problem is ID10Tic not technical. How do you educate an illiterate public on what the difference between alphabetical and lexical sort is in practice?
You can’t, so you engineer around it and call lexical sort alphabetical.