Given that encoded characters must have one and only one General_Category value, it might be too imprecise or arbitrary in some cases. If you ever need more power, it's worth browsing the other character properties Unicode exposes. For example, `Lu` (Uppercase_Letter) only covers some uppercase letters, whereas the `Uppercase` property covers all of them.
---
For anyone that wants to learn more about specific Unicode stuff, the three big data sources are The Core Spec, the Unicode Technical Annexes (UAXs), and the Unicode Character Database itself (the database is a bunch of text files. There's an XML version now as well).
For further reading on this specifically, it might be worth looking at:
Unfortunately, even Properties aren't fully reliable. The Tamil pulli for eg., does not have the Alphabetic property despite being a central part of the Tamil alphabet, pretty much due to historical accident: other Indic languages have a similar looking Virama character that aren't alphabetic, and pulli (despite being a separate, unrelated character) was lumped in with those.
When I tried to raise this in the mailing list and get it rectified, the response I got was pretty much that many properties need language-specific processing anyway, and shouldn't be relied upon fully, and so this wasn't worth fixing.
And where do you get these input strings? Big enough that .split() is not sufficient? Files, and yes sockets support the interface as well with a method call.
What a nonsensical thing to say. You said to call "for line in file" -- you can't do that on a string, even if it originally came from part of a file. Or are you suggesting one should...?
And I said your "generally" was wrong. You were provided general advice, I'm saying it's wrong in general. Do you see me giving numerical quantities anywhere?
No, I didn't misread, input can be self-generated of course. If you're writing a system that's designed like UserInput -> [BlackBox] -> Output, clearly user input won't be auto-generated. But if you factor [BlackBox] into a system like A -> B -> C, A -> D -> C, C -> Output, then each of those arrows will represent an input into the next system that was generated by something our devs wrote. This could be bunch of jsonlines (related to this thread) interpreted as string, a database, some in-memory structure, whatever.
split() without custom delimiter also splits on runs of whitespace, which splitline() also doesn't do (except for \r\n because that combination counts as one line ending):
splitlines() is sometimes nice for adhoc parsing (of well behaved stuff...) because it throws out whitespace-only lines from the resulting list of strings.
#1 use-case of that for me is probably just avoiding the cases where there's a trailing newline character in the output of a command I ran by subprocess.
There's so many super useful things in the Python docs that you never see in the wild. For example, I recently learned that the sqlite3 module has a set_authorizer function that lets you limit the types of statements that can be run / tables that can be accessed.
One thing to note tho is that they take character sets, as long as they encounter characters in the specified set they will keep stripping. Lots of people think if you give it a string it will remove that string.
That feature was added in 3.9 with the addition of `removeprefix` and `removesuffix`.
Sadly,
1. unlike Rust's version they provide no way of knowing whether they stripped things out
2. unlike startswith/endswith they do not take tuples of prefixes/suffixes
I already knew this information, more or less, but I like reading TIL posts like this. It's fun seeing the someone learn new things, and sometimes I pick up something myself, or at least look at it in a new way.
Yeah, don't listen to parent. I like these sorts of articles a lot; its only useless if you assume that everyone interested has also memorized the Python docs fully (which I imagine is zero people). Fun technical tangents are quite fun indeed.
John Yossarian is the protagonist of Joseph Heller’s Catch-22[1], which was my favorite book in high school. Like a lot of people, my handle is a slightly embarrassing memorialization of my younger self :-)
Sometimes value is measured by awareness. I benefited from becoming aware of the behavior because of the article. Yes, it's in the docs, but the docs are not something I would have gone looking to read today.
The value of this article, to me, is that I'd never read the splitlines documentation, so this is a little detail that I just learned thanks to it being linked here.
For more controlled splitting, I really like Unicode named characters classes[0] for more precise splitting and matching tasks.
[0]: https://en.wikipedia.org/wiki/Unicode_character_property#Gen...
Given that encoded characters must have one and only one General_Category value, it might be too imprecise or arbitrary in some cases. If you ever need more power, it's worth browsing the other character properties Unicode exposes. For example, `Lu` (Uppercase_Letter) only covers some uppercase letters, whereas the `Uppercase` property covers all of them.
---
For anyone that wants to learn more about specific Unicode stuff, the three big data sources are The Core Spec, the Unicode Technical Annexes (UAXs), and the Unicode Character Database itself (the database is a bunch of text files. There's an XML version now as well).
For further reading on this specifically, it might be worth looking at:
[Unicode Core Spec - Chapter 4: Character Properties] https://www.unicode.org/versions/Unicode17.0.0/core-spec/cha...
├ [General Category] https://www.unicode.org/versions/Unicode17.0.0/core-spec/cha...
└ [Properties for Text Boundaries] https://www.unicode.org/versions/Unicode17.0.0/core-spec/cha...
[UAX #44 - Unicode Character Database (Technical Report)] https://www.unicode.org/reports/tr44/
├ [General Category Values] https://www.unicode.org/reports/tr44/#General_Category_Value...
└ [Property Definitions] https://www.unicode.org/reports/tr44/#Property_Definitions
And, if you're brave and want to see the data itself (skim through UAX #44 first):
[Unicode Character Database] https://www.unicode.org/Public/17.0.0/ucd/
Unfortunately, even Properties aren't fully reliable. The Tamil pulli for eg., does not have the Alphabetic property despite being a central part of the Tamil alphabet, pretty much due to historical accident: other Indic languages have a similar looking Virama character that aren't alphabetic, and pulli (despite being a separate, unrelated character) was lumped in with those.
When I tried to raise this in the mailing list and get it rectified, the response I got was pretty much that many properties need language-specific processing anyway, and shouldn't be relied upon fully, and so this wasn't worth fixing.
[1] https://util.unicode.org/UnicodeJsps/character.jsp?a=0BCD&B1...
Splitlines is generally not needed. for line in file: is more idiomatic.
Splitlines additionally strips the newline character, functionality which is often (maybe even usually?) desired.
This has been controlled via a boolean parameter since at least 2.0, which as far as I can tell is when this method was added to `str`.
It has similar (but not identical) behaviour though:
I would expect it to have identical behavior.
What if the text is already in a [string] buffer?
StringIO can help, .rstrip() for the sibling comment.
If it's reading from a file, you wouldn't be using splitlines() anyway; you'd use readlines().
For string you’d need to
not every line is read from a file
That's where the generally fits in.
No, because that still assumes files are the general usage.
In my experience, they're not. It's strings.
And where do you get these input strings? Big enough that .split() is not sufficient? Files, and yes sockets support the interface as well with a method call.
> And where do you get these input strings?
From database fields, API calls, JSON values, HTML tag content, function inputs generally, you know -- the normal places.
In my experience, most people aren't dealing directly with files (or streams) most of the time.
Your examples ultimately come from files or sockets, as I mentioned. Especially if big enough to use splitlines on them.
I also used the word generally, so your insistence on quantifying the proportion is a complete waste of time.
What a nonsensical thing to say. You said to call "for line in file" -- you can't do that on a string, even if it originally came from part of a file. Or are you suggesting one should...?
And I said your "generally" was wrong. You were provided general advice, I'm saying it's wrong in general. Do you see me giving numerical quantities anywhere?
They might be programmatically generated, for example.
There are countless sources one can get a string from. Surely you don't think filesystems are the only source of strings?
Input is very rarely auto-generated, though output is.
Surely you haven't misread my comments above to such an extent? Perhaps not familiar with sockets.
No, I didn't misread, input can be self-generated of course. If you're writing a system that's designed like UserInput -> [BlackBox] -> Output, clearly user input won't be auto-generated. But if you factor [BlackBox] into a system like A -> B -> C, A -> D -> C, C -> Output, then each of those arrows will represent an input into the next system that was generated by something our devs wrote. This could be bunch of jsonlines (related to this thread) interpreted as string, a database, some in-memory structure, whatever.
str.split() function does the same:
>>> s = "line1\nline2\rline3\r\nline4\vline5\x1dhello"
>>> s.split() ['line1', 'line2', 'line3', 'line4', 'line5', 'hello']
>>> s.splitlines() ['line1', 'line2', 'line3', 'line4', 'line5', 'hello']
But split() has sep argument to define delimiter according which to split the string.. In which case it provides what you expected to happen:
>>> s.split('\n') ['line1', 'line2\rline3\r', 'line4\x0bline5\x1dhello']
In general you want this:
>>> linesep_splitter = re.compile(r'\n|\r\n?')
>>> linesep_splitter.split(s) ['line1', 'line2', 'line3', 'line4\x0bline5\x1dhello']
In that example str.split() has the same result as str.splitlines(), but it's not in general the same, even without custom delimiter.
str.split() splits on runs of consecutive whitespace, any type of whitespace, including tabs and spaces which splitlines() doesn't do.
split() without custom delimiter also splits on runs of whitespace, which splitline() also doesn't do (except for \r\n because that combination counts as one line ending):splitlines() is sometimes nice for adhoc parsing (of well behaved stuff...) because it throws out whitespace-only lines from the resulting list of strings.
#1 use-case of that for me is probably just avoiding the cases where there's a trailing newline character in the output of a command I ran by subprocess.
Is there a parser ambiguity/confusion vector here?
TIL: Python has a splitlines function
There's so many super useful things in the Python docs that you never see in the wild. For example, I recently learned that the sqlite3 module has a set_authorizer function that lets you limit the types of statements that can be run / tables that can be accessed.
https://www.sqlite.org/c3ref/set_authorizer.html
https://docs.python.org/3/library/sqlite3.html#sqlite3.Conne...
Useful to know for security purposes, surprises like that might cause vulnerabilities..
What, no <br\s*\/?>?
in the same theme, NTLAIL strip(), rstrip(), lstrip() can strip other kinds of characters besides whitespace.
One thing to note tho is that they take character sets, as long as they encounter characters in the specified set they will keep stripping. Lots of people think if you give it a string it will remove that string.
That feature was added in 3.9 with the addition of `removeprefix` and `removesuffix`.
Sadly,
1. unlike Rust's version they provide no way of knowing whether they stripped things out
2. unlike startswith/endswith they do not take tuples of prefixes/suffixes
This article provides no additional value to the splitlines() docs.
The "article" is my TIL mini-blog. What were you expecting besides a "today I learned"?
I already knew this information, more or less, but I like reading TIL posts like this. It's fun seeing the someone learn new things, and sometimes I pick up something myself, or at least look at it in a new way.
Yeah, don't listen to parent. I like these sorts of articles a lot; its only useless if you assume that everyone interested has also memorized the Python docs fully (which I imagine is zero people). Fun technical tangents are quite fun indeed.
What is "yossarian", BTW? I'd gotten confused thinking it was someone else's blog, because I naturally parse that as a surname.
John Yossarian is the protagonist of Joseph Heller’s Catch-22[1], which was my favorite book in high school. Like a lot of people, my handle is a slightly embarrassing memorialization of my younger self :-)
[1]: https://en.wikipedia.org/wiki/Catch-22
Don't be embarrassed, it's a good book (and was my favorite too).
> Like a lot of people, my handle is a slightly embarrassing memorialization of my younger self :-)
... Guilty, actually.
Sometimes value is measured by awareness. I benefited from becoming aware of the behavior because of the article. Yes, it's in the docs, but the docs are not something I would have gone looking to read today.
The value of this article, to me, is that I'd never read the splitlines documentation, so this is a little detail that I just learned thanks to it being linked here.
I've been working with Python for a year or so now, and never knew this. I'm grateful to the author.
For all of us that don't read all documentation for every single method, tool, function or similar, it is, by awarenes, very useful.