Personally prefer it to be in the json module as it just feels more logical. But that's just a personal choice. I didn't mention about dumpf as I see a previous thread that became quite controversial (seems like dumps was a more common usage at that time). Something I forgot to mention : A (non-exact)search for this construct in github (https://github.com/search?q=with+open+%3A+json.load&type=Code) gives 20million+ results. Seems like it's a popular set of statements that people use ... ---- The Nomadic Coder

I like the idea of having these functions, but I don't like overloading the argument to a function with "filename or file-like object" as is common in libraries like Pandas. I think there are a few places the standard library does it, but the separation seems better to me. I don't LOVE the names dumpf() and loadf(), but I don't have an obviously better choice. I guess probably I'd want fromfile() and tofile() as more "modern" names. On Fri, Sep 11, 2020 at 12:17 PM Christopher Barker <pythonchb@gmail.com> wrote:

I"m pretty sure this came up recently, and was pretty much rejected.

Another option would be to have json.dump take a file-like-object or a path-like object -- there's plenty of code out there that does that.

hmm.. maybe that was the version that was rejected.

But I like the idea either way. it always seemed cumbersome to me to write the whole context manager in these kinds of cases.

-CHB

On Fri, Sep 11, 2020 at 2:05 PM The Nomadic Coder < atemysemicolon@gmail.com> wrote:

...

Personally prefer it to be in the json module as it just feels more logical. But that's just a personal choice.

I didn't mention about dumpf as I see a previous thread that became quite controversial (seems like dumps was a more common usage at that time).

Something I forgot to mention : A (non-exact)search for this construct in github (https://github.com/search?q=with+open+%3A+json.load&type=Code) gives 20million+ results. Seems like it's a popular set of statements that people use ...

---- The Nomadic Coder _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/R4CZ2I... Code of Conduct: http://python.org/psf/codeofconduct/

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/4XVZAX... Code of Conduct: http://python.org/psf/codeofconduct/

-- The dead increasingly dominate and strangle both the living and the not-yet born. Vampiric capital and undead corporate persons abuse the lives and control the thoughts of homo faber. Ideas, once born, become abortifacients against new conceptions.

On Fri, 11 Sep 2020 at 19:50, David Mertz <mertz@gnosis.cx> wrote:

I like the idea of having these functions, but I don't like overloading the argument to a function with "filename or file-like object" as is common in libraries like Pandas. I think there are a few places the standard library does it, but the separation seems better to me.

I don't LOVE the names dumpf() and loadf(), but I don't have an obviously better choice. I guess probably I'd want fromfile() and tofile() as more "modern" names.

While separating the methods to avoid method overloading feels much cleaner (and I also think of it that way), w have to look at the other side of it: If .load and .dump are super-charged, people coding with these methods in mind have _one_ less_ thing to worry about: if the method accepts a path or an open file becomes irrelevant. On th othr hand, with two extra methods, anyone coding the samething have one _extra_ thing to worry about, even more things than they have today! Nowadays, one learn that 'ok, so load takes an open file", and that's it. With two new methods it becomes "ok, was it dump or dumpf which took an openfile? Or was it the other way around?" - and that is bad. Parameter overloading has been around from O.O. inception, it is used in other parts of Python, and permissive signatures are part of what made Python so atractive along the years. So, this is "the other side".

On Fri, Sep 11, 2020 at 12:17 PM Christopher Barker <pythonchb@gmail.com> wrote:

...

I"m pretty sure this came up recently, and was pretty much rejected.

Another option would be to have json.dump take a file-like-object or a path-like object -- there's plenty of code out there that does that.

hmm.. maybe that was the version that was rejected.

But I like the idea either way. it always seemed cumbersome to me to write the whole context manager in these kinds of cases.

-CHB

On Fri, Sep 11, 2020 at 2:05 PM The Nomadic Coder < atemysemicolon@gmail.com> wrote:

...

Personally prefer it to be in the json module as it just feels more logical. But that's just a personal choice.

I didn't mention about dumpf as I see a previous thread that became quite controversial (seems like dumps was a more common usage at that time).

Something I forgot to mention : A (non-exact)search for this construct in github (https://github.com/search?q=with+open+%3A+json.load&type=Code) gives 20million+ results. Seems like it's a popular set of statements that people use ...

---- The Nomadic Coder _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/R4CZ2I... Code of Conduct: http://python.org/psf/codeofconduct/

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/4XVZAX... Code of Conduct: http://python.org/psf/codeofconduct/

-- The dead increasingly dominate and strangle both the living and the not-yet born. Vampiric capital and undead corporate persons abuse the lives and control the thoughts of homo faber. Ideas, once born, become abortifacients against new conceptions. _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/WOTLFR... Code of Conduct: http://python.org/psf/codeofconduct/

json.load and json.dump already default to UTF8 and already have parameters for json loading and dumping. json.loads and json.dumps exist only because there was no way to distinguish between a string containing JSON and a file path string. (They probably should've been .loadstr and .dumpstr, but it's too late for that now) TBH, I think it would be great to just have .load and .dump read the file with standard params when a path-like ( hasattr(obj, '__path__') ) is passed, but the suggested disadvantages of this are: - https://docs.python.org/3/library/functions.html#open

The default encoding is platform dependent (whatever locale.getpreferredencoding() returns), but any text encoding supported by Python can be used. See the codecs module for the list of supported encodings.

JSON text SHALL be encoded in UTF-8, UTF-16, or UTF-32. The default encoding is UTF-8, and JSON texts that are encoded in UTF-8 are interoperable in the sense that they will be read successfully by the maximum number of implementations; there are many implementations

- .load and .dump don't default to UTF8? AFAIU, they do default to UTF-8. Do they instead currently default to locale.getpreferredencoding() instead of the JSON spec(s) * encoding= was removed from .loads and was never accepted by json.load or json.dump - .load and .dump would also need to accept an encoding= parameter for non-spec data that don't want to continue handling the file themselves - pickle.load has an encoding= parameter - marshal.load does not have (and probably doesn't need?) an encoding= parameter - What if you need to specify parameters for the file context manager? Accepting a path-like object should not break any existing code: you could always still open and close a file-like yourself. open('file', 'rb') as _file: json.load(_file) - Should we be using open(pth, 'rb') and open(pth, 'wb')? (Binary mode) JSON Specs: - https://tools.ietf.org/html/rfc7159#section-8.1 : that cannot successfully read texts in other encodings (such as UTF-16 and UTF-32). Implementations MUST NOT add a byte order mark to the beginning of a JSON text. In the interests of interoperability, implementations that parse JSON texts MAY ignore the presence of a byte order mark rather than treating it as an error. - https://www.json.org/ > http://www.ecma-international.org/publications/files/ECMA-ST/ECMA-404.pdf (PDF!)

JSON syntax describes a sequence of Unicode code points. JSON also depends on Unicode in the hex numbers used in the \u escapement notation

So, could we just have .load and .dump accept a path-like and an encoding= parameter (because they need to be able to specify UTF-8 / UTF-16 / UTF-32 anyway)? On Tue, Sep 15, 2020 at 3:22 AM Stephen J. Turnbull < turnbull.stephen.fw@u.tsukuba.ac.jp> wrote:

Joao S. O. Bueno writes:

...

If .load and .dump are super-charged, people coding with these methods in mind have _one_ less_ thing to worry about: if the method accepts a path or an open file becomes irrelevant.

But then you either lose the primary benefit of this three line function (defaulting to the UTF-8 encoding to conform to the JSON standard), or you have a situation where what encoding you get can depend on whether you use the name of a file or that file already opened.

I consider that worse because it's precisely the kind of thing that people *don't* worry about and *do* have some difficulty debugging. _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/KO3ZZN... Code of Conduct: http://python.org/psf/codeofconduct/

On Tue, Sep 15, 2020 at 9:09 AM Wes Turner <wes.turner@gmail.com> wrote:

json.load and json.dump already default to UTF8 and already have parameters for json loading and dumping.

yes, of course. json.loads and json.dumps exist only because there was no way to

distinguish between a string containing JSON and a file path string. (They probably should've been .loadstr and .dumpstr, but it's too late for that now)

I think they exist because that was the pickle API from years ago -- though maybe that's why the pickle API had them. Though I think you have it a bit backwards -- you can't pass a path into loads/dumps for that reason. If they were created because that distinction couldn't be made, then load/sump would have accepted a string path back in the day. TBH, I think it would be great to just have .load and .dump read the file

with standard params when a path-like ( hasattr(obj, '__path__') ) is passed, but the suggested disadvantages of this are:

- https://docs.python.org/3/library/functions.html#open

...

The default encoding is platform dependent (whatever locale.getpreferredencoding() returns), but any text encoding supported by Python can be used. See the codecs module for the list of supported encodings.

that's not a reason at all -- the reason is that some folks think overloading a function like this is bad API design. And it's been the way it's been for a long time, so probably better to add a new function(s), rather than extend the API of an existing one.

- .load and .dump don't default to UTF8? AFAIU, they do default to UTF-8. Do they instead currently default to locale.getpreferredencoding() instead of the JSON spec(s) * encoding= was removed from .loads and was never accepted by json.load or json.dump

I think dump defaults to UTF-8. But load is a bit odd (and not that well documented). it appears to accept a file_like object that returns either a string or a byte object from its read() method. If strings, then the decoding is done. if bytes, then I assume that it's using utf-8. This, by the way, should be better documented.

- .load and .dump would also need to accept an encoding= parameter for non-spec data that don't want to continue handling the file themselves - pickle.load has an encoding= parameter

.loads doesn't now, so I don't see why they would need to with the proposed change. You can always encode/decode ahead of time however you want, either in the file-like object or by passing decoded str to .loads/dumps.

- Should we be using open(pth, 'rb') and open(pth, 'wb')? (Binary mode)

no, I think that's clear. in fact, you can't currently dump to a binary file: In [26]: json.dump(obj, open('tiny-enc.json', 'wb')) --------------------------------------------------------------------------- TypeError Traceback (most recent call last) <ipython-input-26-02e9bcd47a3e> in <module> ----> 1 json.dump(obj, open('tiny-enc.json', 'wb')) ~/miniconda3/envs/py3/lib/python3.8/json/__init__.py in dump(obj, fp, skipkeys, ensure_ascii, check_circular, allow_nan, cls, indent, separators, default, sort_keys, **kw) 178 # a debuggability cost 179 for chunk in iterable: --> 180 fp.write(chunk) 181 182 TypeError: a bytes-like object is required, not 'str' That's the beauty of Python 3's text model :-) JSON Specs:

- https://tools.ietf.org/html/rfc7159#section-8.1 :

...

JSON text SHALL be encoded in UTF-8, UTF-16, or UTF-32. The default encoding is UTF-8,

So THAT is interesting. But the current implementation does not directly support anything but UTF-8, and I think it's fine that that still be the case. If anyone is using the other two, it's an esoteric case, and they can encode/decode by hand.

So, could we just have .load and .dump accept a path-like and an encoding= parameter (because they need to be able to specify UTF-8 / UTF-16 / UTF-32 anyway)?

These are separate questions, but I'll say: Yes, it could take a path-like. But I think there was not much support for that in this discussion. No -- there is no need for encoding parameter -- the other two options are rare and can be done by hand. BTW: .dumps() dumps to, well, a string, so it's not assuming any encoding. A user can encode it any way they want when passing it along. This, in fact, is all very Python3 text model compatible -- the encoding/decoding should happen as close to IO as possible. If there were no backward compatibility options, and it were me, I would only use strings in/out of the json module, but I think that ship has sailed. Anyway -- if anyone wants to push for overloading .load()/dump(), rather than making two new loadf() and dumpf() functions, then speak now -- that will take more discussion, and maybe a PEP. -CHB -- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On Tue, Sep 15, 2020 at 5:26 PM Wes Turner <wes.turner@gmail.com> wrote:

On Tue, Sep 15, 2020 at 9:09 AM Wes Turner <wes.turner@gmail.com> wrote:

...

json.load and json.dump already default to UTF8 and already have

...

parameters for json loading and dumping.

so it turns out that loads(), which optionally takes a bytes or bytesarray object tries to determine whether the encoding is UTF-6, UTF-!6 or utf-32 (the ones allowed by the standard) (thanks Guido for the pointer). And load() calls loads(), so it should work with binary mode files as well.

Currently, dump() simply uses the fp passed in, and it doesn't support binary files, so it'll use the encoding the user set (or the default, if not set, which is an issue here) dumps() returns a string, so no encoding there. I think dumpf() should use UTF-8, and that's it. If anyone really wants something else, they can get it by providing an open text file object. loads(), on the other hand, is a bit tricky -- it could allow only UTF-8, but it seems it would be more consistent (and easy to do) to open the file in binary mode and use the existing code to determine the encoding. -CHB

...

The Python JSON implementation should support the full JSON spec (including UTF-8, UTF-16, and UTF-32) and should default to UTF-8.

'turns out it does already, and no one is suggesting changing that. Anyway -- if anyone wants to push for overloading .load()/dump(), rather

...

than making two new loadf() and dumpf() functions, then speak now -- that will take more discussion, and maybe a PEP.

I don't see why one or the other would need a PEP so long as the new functionality is backward-compatible?

iIm just putting my finger in the wind. no need for a PEP if it's simeel and non-controversial, but if even the few folks on this thread don't agree on the API we want, then it's maybe too controversial -- so either more discussion, to come to consensus, or a PEP. Or not -- we can see what the core devs say if/when someone does a bpo / PR. -CHB

...

-CHB

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

-- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

https://docs.python.org/3/library/os.html#os.fspath *__fspath__ On Wed, Sep 16, 2020, 5:53 PM Wes Turner <wes.turner@gmail.com> wrote:

On Wed, Sep 16, 2020, 5:18 PM Christopher Barker <pythonchb@gmail.com> wrote:

...

On Tue, Sep 15, 2020 at 5:26 PM Wes Turner <wes.turner@gmail.com> wrote:

...

On Tue, Sep 15, 2020 at 9:09 AM Wes Turner <wes.turner@gmail.com> wrote:

...

json.load and json.dump already default to UTF8 and already have

...

parameters for json loading and dumping.

so it turns out that loads(), which optionally takes a bytes or bytesarray object tries to determine whether the encoding is UTF-6, UTF-!6 or utf-32 (the ones allowed by the standard) (thanks Guido for the pointer). And load() calls loads(), so it should work with binary mode files as well.

Currently, dump() simply uses the fp passed in, and it doesn't support binary files, so it'll use the encoding the user set (or the default, if not set, which is an issue here) dumps() returns a string, so no encoding there.

So I was not correct: dump does not default to UTF-8 (and does not accept an encoding= parameter)

...

I think dumpf() should use UTF-8, and that's it. If anyone really wants something else, they can get it by providing an open text file object.

Why would we impose UTF-8 when the spec says UTF-8, UTF-16, or UTF-32?

How could this be improved? (I'm on my phone, so)

def dumpf(obj, path, *args, **kwargs): with open(getattr(path, '__path__', path), 'w', encoding=kwargs.get('encoding', 'utf8')) as _file: return dump(_file, *args, **kwargs)

def loadf(obj, path, *args, **kwargs): with open(getattr(path, '__path__', path), encoding=kwargs.get('encoding', 'utf8')) as _file: return load(_file, *args, **kwargs)

...

loads(), on the other hand, is a bit tricky -- it could allow only UTF-8, but it seems it would be more consistent (and easy to do) to open the file in binary mode and use the existing code to determine the encoding.

-CHB

...

...

The Python JSON implementation should support the full JSON spec (including UTF-8, UTF-16, and UTF-32) and should default to UTF-8.

'turns out it does already, and no one is suggesting changing that.

Anyway -- if anyone wants to push for overloading .load()/dump(), rather

...

...

than making two new loadf() and dumpf() functions, then speak now -- that will take more discussion, and maybe a PEP.

I don't see why one or the other would need a PEP so long as the new functionality is backward-compatible?

iIm just putting my finger in the wind. no need for a PEP if it's simeel and non-controversial, but if even the few folks on this thread don't agree on the API we want, then it's maybe too controversial -- so either more discussion, to come to consensus, or a PEP.

Or not -- we can see what the core devs say if/when someone does a bpo / PR.

-CHB

...

...

-CHB

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

Is all of this locale/encoding testing necessary (or even sufficient)? ```python import json import locale import os def get_default_encoding(): """ TODO XXX: ??? """ default_encoding = locale.getdefaultlocale()[1] if default_encoding.startswith("UTF-"): return default_encoding else: return "UTF-8" def dumpf(obj, path, *args, **kwargs): with open( os.fspath(path), "w", encoding=kwargs.pop("encoding", get_default_encoding()), ) as file_: return json.dump(obj, file_, *args, **kwargs) def loadf(path, *args, **kwargs): with open( os.fspath(path), "r", encoding=kwargs.pop("encoding", get_default_encoding()), ) as file_: return json.load(file_, *args, **kwargs) import pathlib import unittest class TestJsonLoadfAndDumpf(unittest.TestCase): def setUp(self): self.locales = ["", "C", "en_US.UTF-8", "japanese"] self.encodings = [None, "UTF-8", "UTF-16", "UTF-32"] data = dict( obj=dict(a=dict(b=[1, 2, 3])), encoding=None, path=pathlib.Path(".") / "test_loadf_and_dumpf.json", ) if os.path.isfile(data["path"]): os.unlink(data["path"]) self.data = data self.previous_locale = locale.getlocale() def tearDown(self): locale.setlocale(locale.LC_ALL, self.previous_locale) def test_get_default_encoding(self): for localestr in self.locales: locale.setlocale(locale.LC_ALL, localestr) output = get_default_encoding() assert output.startswith("UTF-") def test_dumpf_and_loadf(self): data = self.data for localestr in self.locales: locale.setlocale(locale.LC_ALL, localestr) for encoding in self.encodings: dumpf_output = dumpf( data["obj"], data["path"], encoding=encoding ) loadf_output = loadf(data["path"], encoding=encoding) assert loadf_output == data["obj"] ``` On Wed, Sep 16, 2020 at 8:30 PM Christopher Barker <pythonchb@gmail.com> wrote:

On Wed, Sep 16, 2020 at 2:53 PM Wes Turner <wes.turner@gmail.com> wrote:

...

So I was not correct: dump does not default to UTF-8 (and does not accept an encoding= parameter)

...

I think dumpf() should use UTF-8, and that's it. If anyone really wants something else, they can get it by providing an open text file object.

Why would we impose UTF-8 when the spec says UTF-8, UTF-16, or UTF-32?

The idea was that the encoding was one of the motivators to doing this in the first place. But I suppose as long as utf-8 is the default, and only the three "official" ones are allowed, then yeah, we could add an encoding keyword argument.

-CHB

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

Something like this in the docstring?: "In order to support the historical JSON specification and closed ecosystem JSON, it is possible to specify an encoding other than UTF-8." 8.1. Character Encoding

JSON text exchanged between systems that are not part of a closed ecosystem MUST be encoded using UTF-8 [RFC3629]. Previous specifications of JSON have not required the use of UTF-8 when transmitting JSON text. However, the vast majority of JSON- based software implementations have chosen to use the UTF-8 encoding, to the extent that it is the only encoding that achieves interoperability. Implementations MUST NOT add a byte order mark (U+FEFF) to the beginning of a networked-transmitted JSON text. In the interests of interoperability, implementations that parse JSON texts MAY ignore the presence of a byte order mark rather than treating it as an error.

```python import json import os def dumpf(obj, path, *, encoding="UTF-8", **kwargs): with open(os.fspath(path), "w", encoding=encoding) as f: return json.dump(obj, f, **kwargs) def loadf(path, *, encoding="UTF-8", **kwargs): with open(os.fspath(path), "r", encoding=encoding) as f: return json.load(f, **kwargs) import pathlib import unittest class TestJsonLoadfAndDumpf(unittest.TestCase): def setUp(self): self.encodings = [None, "UTF-8", "UTF-16", "UTF-32"] data = dict( obj=dict(a=dict(b=[1, 2, 3])), path=pathlib.Path(".") / "test_loadf_and_dumpf.json", ) if os.path.isfile(data["path"]): os.unlink(data["path"]) self.data = data def test_dumpf_and_loadf(self): data = self.data for encoding in self.encodings: path = f'{data["path"]}.{encoding}.json' dumpf_output = dumpf(data["obj"], path, encoding=encoding) loadf_output = loadf(path, encoding=encoding) assert loadf_output == data["obj"] # $ pip install pytest-cov # $ pytest -v example.py # https://docs.pytest.org/en/stable/parametrize.html # https://docs.pytest.org/en/stable/tmpdir.html import pytest @pytest.mark.parametrize("encoding", [None, "UTF-8", "UTF-16", "UTF-32"]) @pytest.mark.parametrize("obj", [dict(a=dict(b=[1, 2, 3]))]) def test_dumpf_and_loadf(obj, encoding, tmpdir): pth = pathlib.Path(tmpdir) / f"test_loadf_and_dumpf.{encoding}.json" dumpf_output = dumpf(obj, pth, encoding=encoding) loadf_output = loadf(pth, encoding=encoding) assert loadf_output == obj ``` For whoever creates a PR for this: - [ ] add parameter and return type annotations - [ ] copy docstrings from json.load/json.dump and open#encoding - [ ] correctly support the c module implementation (this just does `import json`)? - [ ] keep or drop the encoding tests? On Thu, Sep 17, 2020 at 1:25 AM Christopher Barker <pythonchb@gmail.com> wrote:

Is that suggested code? I don't follow.

But if it is, no. personally, I think ANY use of system settings is a bad idea [*]. But certainly no need to even think about it for JSON.

-CHB

* have we not learned that in the age of the internet the machine the code happens to be running on has nothing to do with the user of the applications' needs? Timezones, encodings, number formats, NOTHING.

On Wed, Sep 16, 2020 at 8:45 PM Wes Turner <wes.turner@gmail.com> wrote:

...

Is all of this locale/encoding testing necessary (or even sufficient)?

```python import json import locale import os

def get_default_encoding(): """ TODO XXX: ??? """ default_encoding = locale.getdefaultlocale()[1] if default_encoding.startswith("UTF-"): return default_encoding else: return "UTF-8"

def dumpf(obj, path, *args, **kwargs): with open( os.fspath(path), "w", encoding=kwargs.pop("encoding", get_default_encoding()), ) as file_: return json.dump(obj, file_, *args, **kwargs)

def loadf(path, *args, **kwargs): with open( os.fspath(path), "r", encoding=kwargs.pop("encoding", get_default_encoding()), ) as file_: return json.load(file_, *args, **kwargs)

import pathlib import unittest

class TestJsonLoadfAndDumpf(unittest.TestCase): def setUp(self): self.locales = ["", "C", "en_US.UTF-8", "japanese"] self.encodings = [None, "UTF-8", "UTF-16", "UTF-32"]

data = dict( obj=dict(a=dict(b=[1, 2, 3])), encoding=None, path=pathlib.Path(".") / "test_loadf_and_dumpf.json", ) if os.path.isfile(data["path"]): os.unlink(data["path"]) self.data = data

self.previous_locale = locale.getlocale()

def tearDown(self): locale.setlocale(locale.LC_ALL, self.previous_locale)

def test_get_default_encoding(self): for localestr in self.locales: locale.setlocale(locale.LC_ALL, localestr) output = get_default_encoding() assert output.startswith("UTF-")

def test_dumpf_and_loadf(self): data = self.data for localestr in self.locales: locale.setlocale(locale.LC_ALL, localestr) for encoding in self.encodings: dumpf_output = dumpf( data["obj"], data["path"], encoding=encoding ) loadf_output = loadf(data["path"], encoding=encoding) assert loadf_output == data["obj"] ```

On Wed, Sep 16, 2020 at 8:30 PM Christopher Barker <pythonchb@gmail.com> wrote:

...

On Wed, Sep 16, 2020 at 2:53 PM Wes Turner <wes.turner@gmail.com> wrote:

...

So I was not correct: dump does not default to UTF-8 (and does not accept an encoding= parameter)

...

I think dumpf() should use UTF-8, and that's it. If anyone really wants something else, they can get it by providing an open text file object.

Why would we impose UTF-8 when the spec says UTF-8, UTF-16, or UTF-32?

The idea was that the encoding was one of the motivators to doing this in the first place. But I suppose as long as utf-8 is the default, and only the three "official" ones are allowed, then yeah, we could add an encoding keyword argument.

-CHB

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On Thu, Sep 17, 2020 at 6:54 AM Wes Turner <wes.turner@gmail.com> wrote:

Why would we impose UTF-8 when the spec says UTF-8, UTF-16, or UTF-32?

Obsolete JSON spec said UTF-8, UTF-16, and UTF-32. Current spec says UTF-8. See https://tools.ietf.org/html/rfc8259#section-8.1 So `dumpf` must use UTF-8, although `loadf` can support UTF-16 and UTF-32 like `loads`.

How could this be improved? (I'm on my phone, so)

def dumpf(obj, path, *args, **kwargs): with open(getattr(path, '__path__', path), 'w', encoding=kwargs.get('encoding', 'utf8')) as _file: return dump(_file, *args, **kwargs)

def loadf(obj, path, *args, **kwargs): with open(getattr(path, '__path__', path), encoding=kwargs.get('encoding', 'utf8')) as _file: return load(_file, *args, **kwargs)

def dumpf(obj, path, *, **kwargs): with open(path, "w", encoding="utf-8") as f: return dump(obj, f, **kwargs) def loadf(obj, path, *, **kwargs): with open(path, "rb") as f: return load(f, **kwargs) Regards, -- Inada Naoki <songofacandy@gmail.com>

What happened to "not every three-line function needs to be a built-in"? This is *literally* a three-line function. And I know the proposal is not to make it a builtin, but still... ISTM down here lies the path to PHP. On Fri, Sep 11, 2020 at 3:16 PM Christopher Barker <pythonchb@gmail.com> wrote:

I"m pretty sure this came up recently, and was pretty much rejected.

Another option would be to have json.dump take a file-like-object or a path-like object -- there's plenty of code out there that does that.

hmm.. maybe that was the version that was rejected.

But I like the idea either way. it always seemed cumbersome to me to write the whole context manager in these kinds of cases.

-CHB

On Fri, Sep 11, 2020 at 2:05 PM The Nomadic Coder < atemysemicolon@gmail.com> wrote:

...

Personally prefer it to be in the json module as it just feels more logical. But that's just a personal choice.

I didn't mention about dumpf as I see a previous thread that became quite controversial (seems like dumps was a more common usage at that time).

Something I forgot to mention : A (non-exact)search for this construct in github (https://github.com/search?q=with+open+%3A+json.load&type=Code) gives 20million+ results. Seems like it's a popular set of statements that people use ...

---- The Nomadic Coder _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/R4CZ2I... Code of Conduct: http://python.org/psf/codeofconduct/

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/4XVZAX... Code of Conduct: http://python.org/psf/codeofconduct/

-- --Guido van Rossum (python.org/~guido) *Pronouns: he/him **(why is my pronoun here?)* <http://feministing.com/2015/02/03/how-using-they-as-a-singular-pronoun-can-c...>

On 11Sep2020 23:09, The Nomadic Coder <atemysemicolon@gmail.com> wrote:

oops, was not aware of "not every three-line function needs to be a built-in"

This came out personal frustration, as I use this 3 line function very, very often, and the whole community does. Still learning-to-navigate what's accepted here and what's not :)

DRY. Why do not people use modules more often (I don't mean the stdlib, I mean extra modules)? My personal solution to this kind of thing is to keep a little library/module of these 3 line things if I use them. Then you just import stuff and use it. 2 trite examples: I've got a cs.lex module with a bunch of little things in there for parsing stuff - identifiers, quoted strings, etc etc; it's on PyPI so using it elsewhere is trivial. Closer to Nomadic's use case, I've got an @strable decorator, thus: @strable def func(f, ...): ... do something with an open file ... It intercepts the first argument: if a str, it opens it as a file (by default, you can provide an arbitrary function for the "open" action). Then the function just has to work with a file (or whatever a str should turn into, domain specific). You could also write: load_json = strable(json.load) and be on your way. This is also in a module (cs.deco), also on PyPI for reuse. Not every three-line function needs to be a built-in, but for the three-line functions _you_ use a lot, write them _once_ and import them from your personal little module-of-three-line-functions. No need to publish to PyPI (extra work) - it's as easy to keep them locally unless you need them elsewhere. But don't rewrite - reuse! Cheers, Cameron Simpson <cs@cskk.id.au>

On Sat, Sep 12, 2020 at 5:29 AM Stephen J. Turnbull < turnbull.stephen.fw@u.tsukuba.ac.jp> wrote:

See Naoki Inada's post for why this might be a good idea even though it's a three-line function. It's not open and shut (for one thing, on most modern systems the system default encoding is already UTF-8),

Actually that fact is the most compelling reason to do this -- since many (most) development systems use UTF-* by default then if you leave the encoding out, your code will work most of the time, pass tests, pass test on the CI, etc, and then barf when it's run on an unusual system. I have to say that I"m not at all sure that I've remembered to set the encoding in my code that reads/writes JSON from files. Time to go check. In fact, I'm getting bytes/str errors in JSON related parts of code I'm porting to py3 right now. This is probably related. -CHB -- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On Fri, Sep 11, 2020 at 4:19 PM David Mertz <mertz@gnosis.cx> wrote:

On Fri, Sep 11, 2020, 12:59 PM Guido van Rossum <guido@python.org> wrote:

...

What happened to "not every three-line function needs to be a built-in"? This is *literally* a three-line function. And I know the proposal is not to make it a builtin, but still... ISTM down here lies the path to PHP.

By the same reasoning, though, if you have dumps(), writing dump() in terms of it is a three-line function. Same in the venerable pickle versions.

Not quite -- if the serialized data is really huge it makes sense to read or write it directly from/to a file, rather than building up a gigantic in-memory buffer as an intermediate structure. (And going in the other direction, if all you have is load/dump, constructing an io.StringIO instance is a fairly awkward bit of idiom. But I'm convinced by Inada's observation that it's easy to have encoding-related bugs -- we should add this in a way that avoids those. And I'm fine with loadf/dumpf as the names, since we already have loads/dumps. -- --Guido van Rossum (python.org/~guido) *Pronouns: he/him **(why is my pronoun here?)* <http://feministing.com/2015/02/03/how-using-they-as-a-singular-pronoun-can-c...>

12.09.20 01:57, Guido van Rossum пише:

What happened to "not every three-line function needs to be a built-in"? This is *literally* a three-line function. And I know the proposal is not to make it a builtin, but still... ISTM down here lies the path to PHP.

Oh, I am very glad that this principle has not yet been forgotten.

On 12/09/20 8:36 pm, Serhiy Storchaka wrote:

it is not hard to write your own helper with interface and defaults that suit you. It will take less time than writing a letter in a mailing list.

Obviously what's needed is an IDE feature such that whenever you write a 3-line function that you haven't used before, it automatically posts a request to python-ideas asking for it to be added to the stdlib. Think of all the time it would save! -- Greg

On Sat, Sep 12, 2020 at 1:40 AM Serhiy Storchaka <storchaka@gmail.com> wrote:

Second, the resulting function would have monstrous interface. open() takes 8 arguments,

well, maybe not -- this would not need to support the entire open() interface: * No one is suggesting getting rid of the current load() function, so if you need to do something less common, you can always open the file yourself. * Many of the options to open() should be always the same when reading/writing json. In fact, that's (IMHO) the most compelling reason to have this built in -- the defaults may be not ideal for JSON, and other specs may be downright wrong. Here they are all eight parameters. mode='r' : this should be either 'r' or 'w' depending on loading or saving, I don't think there's any other (certainly not common) use case for any other mode. buffering=-1 : This seems very reasonable to put in control of the JSON encoder/decoder encoding=None: this is the important one -- json is always UTF-8 yes? errors=None: This also is good ot be in control of the JSON encoder/decoding newline=None: also in control of the encoder closefd=True: I think this is irrelevant to this use case. opener=None: Also does not need to be supported If I have this right, then none of the optional arguments to open() would need to be passed through.

they should be combined. And thank that open() does not accept arbitrary

var-keyword arguments as json.load(), resolving this conflict would be impossible.

see above -- there is no conflict to resolve.

Third, this combination is not so common as you think.

agreed -- but still fairly common. And most of the use cases that aren't file-on-disk are strings anyway. I know I use file-on-disk most of the time I use .load()/dump(), even if I use .loads()/dumps() more frequently. -CHB -- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On Sun, Sep 13, 2020 at 7:58 AM Stephen J. Turnbull < turnbull.stephen.fw@u.tsukuba.ac.jp> wrote:

...

encoding=None: this is the important one -- json is always UTF-8 yes?

Standard JSON is always UTF-8. Nevertheless, I'm quite sure that there's a ton of Japanese in Shift JIS, including some produced by default in Python on Windows. I'll bet the same is true of GBK for Chinese, and maybe even ISO-8859-1 in Europe.

So what should the json lib do with these? It could have an encoding parameter with utf-8 as default. Or it could require that the user open the file themselves if it's not UTF-8. BTW: I noticed that json.loads() takes: Deserialize ``s`` (a ``str``, ``bytes`` or ``bytearray`` instance containing a JSON document) to a Python object. A str is an str (already Unicode, yes?) -- but for bytes, it must be assuming some encoding, presumably UTF-8, but it doesn't seem to have a way to specify one -- so this is already a missing feature. -CHB -- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On 2020-09-13 11:57, Christopher Barker wrote:

On Sun, Sep 13, 2020 at 7:58 AM Stephen J. Turnbull <turnbull.stephen.fw@u.tsukuba.ac.jp <mailto:turnbull.stephen.fw@u.tsukuba.ac.jp>> wrote:

> encoding=None: this is the important one -- json is always UTF-8 yes?

Standard JSON is always UTF-8. Nevertheless, I'm quite sure that there's a ton of Japanese in Shift JIS, including some produced by default in Python on Windows. I'll bet the same is true of GBK for Chinese, and maybe even ISO-8859-1 in Europe.

So what should the json lib do with these? It could have an encoding parameter with utf-8 as default. Or it could require that the user open the file themselves if it's not UTF-8.

BTW: I noticed that json.loads() takes:

Deserialize ``s`` (a ``str``, ``bytes`` or ``bytearray`` instance containing a JSON document) to a Python object.

A str is an str (already Unicode, yes?) -- but for bytes, it must be assuming some encoding, presumably UTF-8, but it doesn't seem to have a way to specify one -- so this is already a missing feature.

It's not a missing feature, because the JSON spec requires UTF-8. If it's not UTF-8, it's invalid JSON. If a user wants to handle a file that looks sort of like JSON but technically isn't because it's not UTF-8, it's on the user to first convert the file to UTF-8 before bringing JSON into the picture. -- Brendan Barnwell "Do not follow where the path may lead. Go, instead, where there is no path, and leave a trail." --author unknown

Yes, that is a design flaw in the stdlib. There ought to be an opt-in switch for accepting/producing those special values, not the current opt-out for strictness... And the misnamed parameter is 'allow_nan' whereas it also configures 'Infinity'. On Sun, Sep 13, 2020, 3:16 PM Matthias Bussonnier < bussonniermatthias@gmail.com> wrote:

Oh, well, but stdlib json already emit (and parse) invalid json by default...

$ ipython Python 3.8.3 (default, May 19 2020, 13:54:14) Type 'copyright', 'credits' or 'license' for more information IPython 7.15.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import json

In [2]: json.dumps({'a':float('nan')}) Out[2]: '{"a": NaN}'

In [3]: json.loads(json.dumps({'a':float('nan')})) Out[3]: {'a': nan}

NaN is not allowed in json spec, so this is wrong, though you can still do it... Maybe it's one of those cases where practicality beats purity. -- M

On Sun, 13 Sep 2020 at 12:31, Brendan Barnwell <brenbarn@brenbarn.net> wrote:

...

On 2020-09-13 11:57, Christopher Barker wrote:

...

On Sun, Sep 13, 2020 at 7:58 AM Stephen J. Turnbull <turnbull.stephen.fw@u.tsukuba.ac.jp <mailto:turnbull.stephen.fw@u.tsukuba.ac.jp>> wrote:

> encoding=None: this is the important one -- json is always

UTF-8

...

...

yes?

Standard JSON is always UTF-8. Nevertheless, I'm quite sure that there's a ton of Japanese in Shift JIS, including some produced by default in Python on Windows. I'll bet the same is true of GBK for Chinese, and maybe even ISO-8859-1 in Europe.

So what should the json lib do with these? It could have an encoding parameter with utf-8 as default. Or it could require that the user open the file themselves if it's not UTF-8.

BTW: I noticed that json.loads() takes:

Deserialize ``s`` (a ``str``, ``bytes`` or ``bytearray`` instance containing a JSON document) to a Python object.

A str is an str (already Unicode, yes?) -- but for bytes, it must be assuming some encoding, presumably UTF-8, but it doesn't seem to have a way to specify one -- so this is already a missing feature.

It's not a missing feature, because the JSON spec requires UTF-8. If it's not UTF-8, it's invalid JSON. If a user wants to handle a file that looks sort of like JSON but technically isn't because it's not UTF-8, it's on the user to first convert the file to UTF-8 before bringing JSON into the picture.

-- Brendan Barnwell "Do not follow where the path may lead. Go, instead, where there is no path, and leave a trail." --author unknown _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/6LLNXB... Code of Conduct: http://python.org/psf/codeofconduct/

---

Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/R36SES... Code of Conduct: http://python.org/psf/codeofconduct/

From "[Python-ideas] JSON encoder protocol (was Re: adding support for a "raw output" in JSON serializer)" https://mail.python.org/archives/list/python-ideas@python.org/message/5C4UHZ...

- There's JSON5; which supports comments, trailing commas, IEEE 754 ±Infinity and NaN, [...] https://json5.org/

On Sun, Sep 13, 2020, 9:53 PM David Mertz <mertz@gnosis.cx> wrote:

Yes, that is a design flaw in the stdlib. There ought to be an opt-in switch for accepting/producing those special values, not the current opt-out for strictness... And the misnamed parameter is 'allow_nan' whereas it also configures 'Infinity'.

On Sun, Sep 13, 2020, 3:16 PM Matthias Bussonnier < bussonniermatthias@gmail.com> wrote:

...

Oh, well, but stdlib json already emit (and parse) invalid json by default...

$ ipython Python 3.8.3 (default, May 19 2020, 13:54:14) Type 'copyright', 'credits' or 'license' for more information IPython 7.15.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import json

In [2]: json.dumps({'a':float('nan')}) Out[2]: '{"a": NaN}'

In [3]: json.loads(json.dumps({'a':float('nan')})) Out[3]: {'a': nan}

NaN is not allowed in json spec, so this is wrong, though you can still do it... Maybe it's one of those cases where practicality beats purity. -- M

On Sun, 13 Sep 2020 at 12:31, Brendan Barnwell <brenbarn@brenbarn.net> wrote:

...

On 2020-09-13 11:57, Christopher Barker wrote:

...

On Sun, Sep 13, 2020 at 7:58 AM Stephen J. Turnbull <turnbull.stephen.fw@u.tsukuba.ac.jp <mailto:turnbull.stephen.fw@u.tsukuba.ac.jp>> wrote:

> encoding=None: this is the important one -- json is always

UTF-8

...

...

yes?

Standard JSON is always UTF-8. Nevertheless, I'm quite sure that there's a ton of Japanese in Shift JIS, including some produced by default in Python on Windows. I'll bet the same is true of GBK

for

...

Chinese, and maybe even ISO-8859-1 in Europe.

So what should the json lib do with these? It could have an encoding parameter with utf-8 as default. Or it could require that the user

open

...

the file themselves if it's not UTF-8.

BTW: I noticed that json.loads() takes:

Deserialize ``s`` (a ``str``, ``bytes`` or ``bytearray`` instance containing a JSON document) to a Python object.

A str is an str (already Unicode, yes?) -- but for bytes, it must be assuming some encoding, presumably UTF-8, but it doesn't seem to have a way to specify one -- so this is already a missing feature.

It's not a missing feature, because the JSON spec requires UTF-8. If it's not UTF-8, it's invalid JSON. If a user wants to handle a file that looks sort of like JSON but technically isn't because it's not UTF-8, it's on the user to first convert the file to UTF-8 before bringing JSON into the picture.

-- Brendan Barnwell "Do not follow where the path may lead. Go, instead, where there is no path, and leave a trail." --author unknown _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/6LLNXB... Code of Conduct: http://python.org/psf/codeofconduct/

---

Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/R36SES... Code of Conduct: http://python.org/psf/codeofconduct/

_______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/RYKDII... Code of Conduct: http://python.org/psf/codeofconduct/

There seems to be a fair bit of support for this idea. Will it need a PEP ? -CHB On Mon, Sep 14, 2020 at 9:20 AM Stephen J. Turnbull < turnbull.stephen.fw@u.tsukuba.ac.jp> wrote:

Christopher Barker writes:

...

On Sun, Sep 13, 2020 at 7:58 AM Stephen J. Turnbull <

...

turnbull.stephen.fw@u.tsukuba.ac.jp> wrote:

...

...

...

...

encoding=None: this is the important one -- json is always UTF-8 yes?

...

...

...

...

Standard JSON is always UTF-8. Nevertheless, I'm quite sure that

...

...

there's a ton of Japanese in Shift JIS, including some produced by

...

...

default in Python on Windows. I'll bet the same is true of GBK for

...

...

Chinese, and maybe even ISO-8859-1 in Europe.

...

...

So what should the json lib do with these?

Well, I'm a mail guy from way back, so I'm with Mr. Postol: be

libertine in what you accept, puritan in what you emit. I think given

the current architecture of json, dump and load are fine as is, dump

should be discourage (but not removed!) in favor of dumpf, and dumpf

and loadf should provide no option but UTF-8.

I just wanted to point out that it's very likely that there's a lot of

"JSON-like" data out there, and probably a lot of "unwritten

protocols" that expect it. While nobody has proposed removing dump

and load, I don't want them deprecated or discouraged for the purpose

of dealing with "JSON-like" data, especially not load.

-- Christopher Barker, PhD

Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On Mon, Sep 14, 2020 at 10:00 AM Chris Angelico <rosuav@gmail.com> wrote:

On Tue, Sep 15, 2020 at 2:41 AM Christopher Barker <pythonchb@gmail.com> wrote:

...

There seems to be a fair bit of support for this idea.

Will it need a PEP ?

I think not; unless there are API details to be hashed out or counterarguments to be rebuffed, I think this is fairly simple and non-controversial.

agreed -- at least the simple part. The only API detail I"ve seen is whether to make json.load() take a "path-like or file-like" object, or to have separate functions. And I think the consensus is to have separate functions. And to name them ``loadf()`` and ``savef``. The other potential API issue is whether to support arbitrary encdoings -- but I think the consensus there is also no. So yes -- pretty simple.

Can you create a PR directly? If not, create a bugs.python.org issue

to track it.

not me -- I'm having enough trouble finding time to srite the inf/nan PEP. But I think the OP offered to do so.

Either way, it's up to the core devs to decide, but it shouldn't need all the overhead of a full PEP.

And if it does, the core devs can tell us then. -CHB -- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

Well, I didn't read the entire discussion... but I wrote in unsuspicious times a stupid little module, msutils, with two stupid little functions, jsonLoad and jsonDump: https://github.com/Marco-Sulla/msutils/blob/master/msutils/jsonutil.py#L20

On 14/09/2020 17:36, Christopher Barker wrote:

There seems to be a fair bit of support for this idea.

Will it need a PEP ?

-CHB

If I've understood correctly (far from certain) the existing json.dumps and json.loads functions are permissive (allow some constructions that are not part of the JSON spec) but the proposed new functions will be strict. To minimise possible confusion, I think that the documentation (both the docstrings and the online docs) should be **very clear** about this. E.g. loads:     ...     loads accepts blah-blah-blah.  This is different from loadf which only accepts strict JSON. loadf:     ...     loadf only accepts strict JSON.  This is different from loads which blah-blah-blah Etc. Rob Cliffe

I believe Sergie already suggested pickle and marshall, and I guess we can add plistlib to those. Personally, I'm not so sure it should be added to all these. I see why the same API was used for all of them, but they really are fairly different beasts. So if they have a function with the same purpose, it should have the same name, but that doesn't mean that all these modules need to have all the functions. On the other hand, the fact that we might be adding two new functions to four different modules is, in my mind, andn argument for overloading the existing dump() / load() instead. a lot less API churn. -CHB On Wed, Sep 16, 2020 at 5:10 PM Chris Angelico <rosuav@gmail.com> wrote:

On Thu, Sep 17, 2020 at 9:53 AM <lammenspaolo@gmail.com> wrote:

...

Maybe unrelated, but the same goes for `pickle.load` and `pickle.dump`.

For consistencies, any changes made to `json.load` and `json.dump` (e.g. adding `json.loadf` and `json.dumpf` or accepting a path like as argument) should be also applied equivalently to `pickle.load` and `pickle.dump`.

...

Off the top of my head, I can't think of any more places in the standard

library with the same parallel structure.

...

marshal is the other one in that set, and a quick 'git grep' shows that plistlib also has that API. The xmlrpc.client module also has dumps/loads, but not dump/load.

ChrisA _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/AWJNAL... Code of Conduct: http://python.org/psf/codeofconduct/

-- Christopher Barker, PhD Python Language Consulting - Teaching - Scientific Software Development - Desktop GUI and Web Development - wxPython, numpy, scipy, Cython

On Fri, Sep 18, 2020 at 12:07 AM Paolo Lammens <lammenspaolo@gmail.com> wrote:

Besides, I don't understand what the downside of overloading is, apart from purism (?).

I am one of who are conservative about overloading. I agree this is purism, but I want to explain behind of this purism. In statically, and nominal typed language, overloading is simple and clear because only one type is chosen by compiler. On the other hand, compiler or VM can not choose single type in duck-typed (or structural typed) languages. For example, * str subtype can implement read/write method. It is both of PathLike and file-like. * File subtype can implement `.__fspath__`. It is both of PathLike and File. Of course, statically typed languages like Java allow implementing multiple interfaces. But Java programmer must choose one interface explicitly when it is ambiguous. So it is explicit what type is used in overloading. On the other hand, in case of Python, there are no compiler/VM support for overloading, because Python is duck-typed language. * `load(f, ...)` uses `f.read()` * `dump(f, ...)` uses `f.write()` * `loadf(path, ..)` and `dumpf(path, ...)` uses `open(path, ...)` This is so natural design for duck-typed language. Regards, -- Inada Naoki <songofacandy@gmail.com>

On Fri, Sep 18, 2020 at 1:05 PM Christopher Barker <pythonchb@gmail.com> wrote:

I see the issue here, but isn't that inherent in duck typing? It's not specific to overloading. In fact the current implementation of json.load() simply calls the .read() in the object passed in. If it does not have a read method, you get an AttributeError. IF someone where to pass in a string subclass with a read method, and that method returned the right thing, it would "just work". Here's an example:

Yes - it would indeed just work, and that's because the function *name* specifies whether it's going to read from a file object or open a path name. There's no ambiguity; if you call load() on something that's both a path and a file, it'll work, and if you call loadf() on something that's both a path and a file, it'll also just work. But if you try to make those into one and the same function, how's it going to pick from the two options? ChrisA

I understand the theory here: but practicality beats purity -- do we really imagine that folks will create these path-like and file-like objects, and then pass them into json.load() and get upset when it behaves oddly?

And this really doesn't feel that different from the fact that you can create an object with a .read() method that does something different, and it will fail then, too.

+1, emphasis on the *practicality beats purity*. The whole reason this change was proposed is practicality. Adding two separate functions doesn't solve that problem. (Or, to be more precise, it does solve it, but the complexity it introduces directly counteracts any practicality it introduces.)

- inline the existing contents of _load by indenting within the conditional - just add the new code at the top of the function

Yes, this is what I was referring to; I'd choose one of these options. I'm not so sure about dispatch because it is solely based on type and isn't very flexible; if in the future any change is made to the definition of path-like, or another addition is made to `json.load`/`json.dump`, or whatever, the type-based dispatch option is more likely to need more changes. On Fri, 18 Sep 2020 at 04:22, Wes Turner <wes.turner@gmail.com> wrote:

There's a small amount of overhead to overloading: - 2x conditionals - 1x function call overhead (additional stack frame)

Instead of this approach we could either - inline the existing contents of _load by indenting within the conditional - just add the new code at the top of the function - use multiple dispatch

```python import os import json import pathlib

def test_load(n): print(('n', n, load)) with open('test.json','w') as f: json.dump(True, f) assert load('test.json') assert load(pathlib.Path('test.json')) with open('test.json', 'r') as f: assert load(f)

def _load(file_, **kwargs): # ... rename existing json.load to json._load ... print(('load', file_, kwargs)) return True

def load(file_, *, encoding="UTF-8", **kwargs): if isinstance(file_, str) or hasattr(file_, "__fspath__"): with open(os.fspath(file_), "r", encoding=encoding) as f: return _load(f, **kwargs) else: return _load(file_, **kwargs)

test_load(1)

def load(file_, *, encoding="UTF-8", **kwargs): if isinstance(file_, str) or hasattr(file_, "__fspath__"): with open(os.fspath(file_), "r", encoding=encoding) as f: return _load(f, **kwargs) return _load(file_, **kwargs) # (or inline the existing json.load)

test_load(2)

## singledispatch # https://docs.python.org/3/library/functools.html?highlight=dispatch#functool...

from functools import singledispatch

@singledispatch def load(file_, **kwargs): return _load(file_, **kwargs)

@load.register(str) def load_str(file_: str, *, encoding='UTF-8', **kwargs): with open(file_, "r", encoding=encoding) as f: return _load(f, **kwargs)

@load.register(os.PathLike) def load_pathlike(file_: os.PathLike, *, encoding='UTF-8', **kwargs): with open(os.fspath(file_), "r", encoding=encoding) as f: return _load(f, **kwargs)

test_load(3) ```

On Thu, Sep 17, 2020 at 10:08 PM Inada Naoki <songofacandy@gmail.com> wrote:

...

On Fri, Sep 18, 2020 at 12:07 AM Paolo Lammens <lammenspaolo@gmail.com> wrote:

...

Besides, I don't understand what the downside of overloading is, apart

from purism (?).

I am one of who are conservative about overloading. I agree this is purism, but I want to explain behind of this purism.

In statically, and nominal typed language, overloading is simple and clear because only one type is chosen by compiler. On the other hand, compiler or VM can not choose single type in duck-typed (or structural typed) languages.

For example,

* str subtype can implement read/write method. It is both of PathLike and file-like. * File subtype can implement `.__fspath__`. It is both of PathLike and File.

Of course, statically typed languages like Java allow implementing multiple interfaces. But Java programmer must choose one interface explicitly when it is ambiguous. So it is explicit what type is used in overloading.

On the other hand, in case of Python, there are no compiler/VM support for overloading, because Python is duck-typed language.

* `load(f, ...)` uses `f.read()` * `dump(f, ...)` uses `f.write()` * `loadf(path, ..)` and `dumpf(path, ...)` uses `open(path, ...)`

This is so natural design for duck-typed language.

Regards, -- Inada Naoki <songofacandy@gmail.com> _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/ZZOHJI... Code of Conduct: http://python.org/psf/codeofconduct/

https://github.com/python/cpython/blame/master/Lib/unittest/test/testmock/te... On Mon, Sep 28, 2020 at 12:00 PM Wes Turner <wes.turner@gmail.com> wrote:

`my_dict = (json.load(f) with open(filename.json, "r") as f)`

Would that be called a generator expression / comprehension context manager?

https://docs.python.org/3/reference/datamodel.html#context-managers

https://docs.python.org/3/library/contextlib.html

PEP 343 added the "with" statement.

Tests that would need to be extended / that may be useful references:

https://github.com/python/cpython/blob/master/Lib/test/test_with.py

https://github.com/python/cpython/blob/master/Lib/test/test_contextlib.py

https://github.com/python/cpython/blob/master/Lib/test/test_contextlib_async...

https://github.com/python/cpython/blob/master/Lib/test/test_grammar.py#L1701 test_with_statement

https://github.com/python/cpython/blob/master/Lib/test/test_grammar.py#L1864 test_async_with

https://github.com/python/cpython/blob/master/Grammar/python.gram

Would there be any new scope issues?

On Mon, Sep 28, 2020, 11:12 AM Eric Wieser <wieser.eric+numpy@gmail.com> wrote:

...

This reminds me of a previous proposal I can't remember if I hit the list with, allowing

with open(filename.json, "r") as f: my_dict = json.load(f)

to be spelt as a single expression:

my_dict = (json.load(f) with open(filename.json, "r") as f)

Obviously this would be more useful in comprehensions, but it might encourage people not to lazily write `json.load(open("filename.json", "r"))` because they want an expression, and end up leaving files open.

Eric _______________________________________________ Python-ideas mailing list -- python-ideas@python.org To unsubscribe send an email to python-ideas-leave@python.org https://mail.python.org/mailman3/lists/python-ideas.python.org/ Message archived at https://mail.python.org/archives/list/python-ideas@python.org/message/QR233S... Code of Conduct: http://python.org/psf/codeofconduct/