Add methods to `pathlib.Path`: `write_text`, `read_text`, `write_bytes`, `read_bytes`

[cool-RR]

BPO	20218
Nosy	@birkenfeld, @pitrou, @cool-RR
Files	pathlib.readwrite3.patch: minimal pathlib read/write functions with tests pathlib.readwrite4.patch pathlib.readwrite3_with_exclusive2.patch: minimal pathlib read/write functions with tests (exclusive option added, typo fixed) patch.patch pathlib_readwrite_v5.patch pathlib_readwrite_v6.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = <Date 2014-10-01.17:14:27.523>
created_at = <Date 2014-01-10.19:04:53.510>
labels = ['type-feature', 'library']
title = 'Add methods to `pathlib.Path`: `write_text`, `read_text`, `write_bytes`, `read_bytes`'
updated_at = <Date 2014-10-01.17:14:43.718>
user = 'https://github.com/cool-RR'

bugs.python.org fields:

activity = <Date 2014-10-01.17:14:43.718>
actor = 'pitrou'
assignee = 'none'
closed = True
closed_date = <Date 2014-10-01.17:14:27.523>
closer = 'python-dev'
components = ['Library (Lib)']
creation = <Date 2014-01-10.19:04:53.510>
creator = 'cool-RR'
dependencies = []
files = ['33670', '33729', '33786', '33840', '36761', '36762']
hgrepos = []
issue_num = 20218
keywords = ['patch']
message_count = 50.0
messages = ['207874', '207952', '208604', '208605', '208625', '208635', '208636', '208639', '208640', '208744', '208785', '208878', '208879', '208889', '208924', '208930', '208931', '208940', '208947', '208949', '208951', '208991', '208992', '208999', '209002', '209017', '209029', '209053', '209326', '209584', '209594', '209606', '209832', '210194', '210341', '211925', '214141', '214142', '214144', '214151', '214228', '214289', '227898', '227900', '227913', '227914', '227926', '227947', '228091', '228092']
nosy_count = 6.0
nosy_names = ['georg.brandl', 'pitrou', 'matthiastroffaes', 'cool-RR', 'python-dev', 'cjwelborn']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue20218'
versions = ['Python 3.5']

[cool-RR]

I'd really like to have methods pathlib.Path.write and pathlib.Path.read. Untested implementation:

def read(self, binary=False):
    with self.open('br' is binary else 'r') as file:
        return file.read()

def write(self, data. binary=False):
    with self.open('bw' is binary else 'w') as file:
        file.write(data)

This will be super useful to me. Many files actions are one liners like that, and avoiding putting the with clause in user code would be wonderful.

Antoine suggests that binary shouldn't be an argument, that there should be separate methods for reading/writing text and binary contents, and that the text one would require passing in encoding and other parameters. I'll be happy to add these to the implementation and create a patch, once people can define which parameters should be used.

[cool-RR]

(Replace is with if in my code sample, typo.)

[cjwelborn]

Isn't this something you could do yourself?

import pathlib
def pathread(self, binary=False):
    with self.open('br'if binary else 'r') as fread:
        return fread.read()

def pathwrite(self, data, mode='w'):
    with self.open(mode) as fwrite:
        fwrite.write(data)

		
pathlib.Path.read = pathread
pathlib.Path.write = pathwrite
p = pathlib.Path('/mydir/example.txt')
p.read()
# 'Content from path.\n'

p.write('I am appending.\n', mode='a')
p.read()
# 'Content from path.\nI am appending.\n'

...and what about:
"There should be one-- and preferably only one --obvious way to do it."

[cjwelborn]

I was referring to writing helper functions like pathread(mypath),
though the monkeypatch version (p.read()) does work (it did when I tested it).

[serhiy-storchaka]

Not every two line function are worth to be added to the stdlib.

[cool-RR]

Christopher and Serhiy, I would appreciate if you could kindly explain why your arguments, while applying to my suggestions, do not apply to the following functions:

• Path.stat
• Path.owner
• Path.group
• Path.open
• Path.chmod
• Path.lchmod
• Path.unlink
• Path.rmdir
• Path.lstat
• Path.rename
• Path.replace
• Path.symlink_to

I am quite sick of all those lofty principles being quoted when I want to add something to Python, yet I see Python features that break them all the time.

[pitrou]

FWIW, I agree that shortcuts to easily create or read entire files are useful. Other path classes (such as Twisted's) often have them.

[serhiy-storchaka]

Christopher and Serhiy, I would appreciate if you could kindly explain why your arguments, while applying to my suggestions, do not apply to the following functions:

1. Path.stat() wraps only one function, while Path.read() wraps two functions. It's signature should be a sum of open() and read() signatures: Path.read(encoding, errors, newline, size) (I have omitted some open's parameters). This is a little cumbersome.

2. os.stat() requires import os, while open() is builtin.

3. If add Path.read(), what about Path.readlines() and Path.readline()?

4. I believe Path.stat() will be used much more often than Path.read().

5. Path.stat() corresponds to low-level os.stat(), but for low-level os.read() the high-level corresponding is FileIO.read(). And this corresponding is much more universal and useful.

[cool-RR]

Serhiy:

Your arguments 1 and 2 are pretty weak. (So what if an import is required? It's still 2 lines. I thought that "Not every two line function are worth to be added to the stdlib.")

Regarding stat being used much more often than read: I disagree. I've done whole-file reads much more often than I've done stat. Different people might have different habits, but I definitely think that reading an entire file is a common operation with files. And we can all agree that other trivial Path method like lstat are much less common than both stat and whole-file read, yet they've earned their place in the stdlib, while we are still arguing whether Path.read and path.write should have a place too.

Regarding Path.readline and Path.readlines: I don't have an opinion on whether they should be implemented or not, since I rarely use them. But I definitely think that they have no bearing on the decision of whether to include Path.read and Path.write or not.

Your argument 5 also looks weak to me. It looks to me like you're looking for arguments.

[cjwelborn]

Ram Rachum: If I actually had a say in the matter, I wouldn't be totally against it. (so far I only lurk the lists/tracker looking for things to work on.) I was just offering you a solution that you can use _today_.

I do think that if something like this was added, then you would have to go ahead and add the other read/write functions because the next question would be 'how come pathlib has no readlines()?'.

Also, it would definitely need more than 'mode' or 'binary' for args. The encoding option needs to be taken into account. People need to have a choice in what encoding to use even with a little read()/write().

[cool-RR]

Patch attached. Is this good?

[cjwelborn]

Ram Rachum (and to whom it may concern):

This patch was missing the size argument for readlines(), and it did not match the overall style of pathlib.py. (using ''' instead of """, and other docstring style). It also clobbered the builtin 'file'.

I've attached another patch that tries very hard to match that 'style', and wrote tests that try to match what the original 'io' functions test for.

This is my first patch for python-dev, I am attaching it in the hope it will help resolve the issue, or generate more discussion about it.

I ran both the individual test_pathlib tests, and the entire test suite. Also, the patchcheck sanity test.

[cjwelborn]

Sorry Antoine, I should've done my homework. I didn't realize 'to whom it may concern' is actually you. :)