Compile code with Mypyc #165

ofek · 2022-01-05T04:35:21Z

Hello again! Let me know what you think 😄

This PR compiles code with Mypyc. To do so, I've switched the build system to hatchling in order to use the hatch-mypyc build hook plugin. Hatchling is the minimal build backend of Hatch (v1 rewrite of CLI being released next week). You can think of it as the same distinction between flit-core & flit or poetry-core & poetry, with hatchling being more similar to flit-core in that only accepted standards are used e.g. project metadata format (PEP 621), dependency specification (PEP 631), etc.

I added binary wheels to the release workflow using the now ubiquitous cibuildwheel. Here is an example of the files produced: https://pypi.org/project/hatch-showcase/#files

The only thing I couldn't preserve was the hack where we change the __module__ of TOMLDecodeError.

On my Windows machine, tox -e benchmark shows a 2-3x performance improvement.

ofek · 2022-01-05T05:17:57Z

pyproject.toml

 description = "A lil' TOML parser"
 authors = [
    { name = "Taneli Hukkinen", email = "hukkin@users.noreply.github.com" },
 ]
-license = { file = "LICENSE" }
+license = "MIT"


Supports https://www.python.org/dev/peps/pep-0639/

File automatically included

ofek · 2022-01-05T05:19:24Z

.github/workflows/tests.yaml

-    - name: Build and check
-      run: |
-        rm -rf dist/ && python -m build
-        twine check --strict dist/*


Does this still https://github.com/pypa/gh-action-pypi-publish#disabling-metadata-verification

ofek · 2022-01-05T05:20:41Z

src/tomli/_parser.py

@@ -125,10 +125,10 @@ class Flags:
    """Flags that map to parsed keys/namespaces."""

    # Marks an immutable namespace (inline array or inline table).
-    FROZEN = 0
+    FROZEN: ClassVar = 0


tomli\_parser.py:279: error: Cannot access instance attribute "EXPLICIT_NEST" through class object tomli\_parser.py:279: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:279: error: Cannot access instance attribute "FROZEN" through class object tomli\_parser.py:281: error: Cannot access instance attribute "EXPLICIT_NEST" through class object tomli\_parser.py:281: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:297: error: Cannot access instance attribute "FROZEN" through class object tomli\_parser.py:297: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:302: error: Cannot access instance attribute "EXPLICIT_NEST" through class object tomli\_parser.py:302: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:323: error: Cannot access instance attribute "EXPLICIT_NEST" through class object tomli\_parser.py:323: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:327: error: Cannot access instance attribute "EXPLICIT_NEST" through class object tomli\_parser.py:327: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:329: error: Cannot access instance attribute "FROZEN" through class object tomli\_parser.py:329: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:342: error: Cannot access instance attribute "FROZEN" through class object tomli\_parser.py:342: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:437: error: Cannot access instance attribute "FROZEN" through class object tomli\_parser.py:437: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute) tomli\_parser.py:453: error: Cannot access instance attribute "FROZEN" through class object tomli\_parser.py:453: note: (Hint: Use "x: Final = ..." or "x: ClassVar = ..." to define a class attribute)

ofek · 2022-01-05T05:21:41Z

src/tomli/_parser.py

-            cont = cont[key_stem]
-            return flag in cont["flags"] or flag in cont["recursive_flags"]


TypeError: dict object expected; got set

ofek · 2022-01-05T05:22:28Z

src/tomli/_parser.py

-    relative_path_cont_keys = (header + key[:i] for i in range(1, len(key)))
-    for cont_key in relative_path_cont_keys:


tomli\_parser.py:320: warning: Treating generator comprehension as list

ofek · 2022-01-05T05:24:09Z

src/tomli/__init__.py

-# Pretend this exception was created here.
-TOMLDecodeError.__module__ = __name__


tomli\__init__.py:9: error: Only class variables defined as ClassVar can be assigned to

hukkin · 2022-01-05T09:05:43Z

Hey, wow, thanks for the PR!

On my Windows machine, tox -e benchmark shows a 2-3x performance improvement.

Niiice! This should mean almost the same performance as the C++ and Rust extended pytomlpp and rtoml. With nothing but Python code. Actually pretty amazing!

Let me know what you think

Unfortunately, I think it may be a bad move to merge this into Tomli right now:

Tomli has major role in PEP517 based Python bootstrapping right now. Getting that right has involved considering different build-backends, vendoring options, bootstrap processes where PYTHONPATH is manipulated etc. There's many tickets about the struggles we've faced in Tomli's, Flit's and I think even pip's issue tracker. Now that we've got a set-up that seems to satisfy repackagers, I'm hesitant to change anything.
I believe this would mean a small bit of extra work for repackagers, adapting their processes, which I'd like to spare them from right now.

If #141 goes as I hope, it could be very interesting to return to this in a few years, when Tomli's role in the bootstrapping processes is replaced by a stdlib parser.

And yeah thanks for hatch-mypyc in general! Some months ago I was looking for a way to mypycify mdformat but there wasn't a way that allowed me to keep pyproject.toml, PEP621 metadata etc. I'll probably try these changes out in that repository soon.

ofek · 2022-01-06T08:01:42Z

@hukkin Fixed 😄 I added https://ofek.dev/hatch/latest/config/build/#conditional-execution and now the build hook is disabled by default

hukkin · 2022-01-06T09:22:30Z

Ah sorry for not being clear enough. I think it may not be a good idea to switch build backend right now.

ofek · 2022-01-06T15:27:32Z

Oh I see. Any concern in particular or just not the right time?

hukkin · 2022-01-06T15:40:49Z

My concern is PEP517 bootstrapping issues of distro packagers, where Tomli and its build backend play a big part. The lack of a standard library TOML parser has been a pain for distro packagers, and my understanding is now that flit started vendoring Tomli, they seem somewhat content with the current state. I wouldn't want to disrupt them right now.

Tomli is very widely repackaged now, so changing the build backend means a small bit of extra busywork for a fairly large amount of people.

As soon as a TOML parser moves into the standard library the situation should be very different.

ofek · 2022-01-06T16:35:04Z

Okay, feel free to close this out!

Just so I understand though, can you give an example of what a distro packager would need to do if this was merged? I want hatchling to be as easy to use or package as possible. Is it just a matter of adding the ability to e.g. apt install python-hatchling?

hukkin · 2022-01-06T16:48:28Z

I'm not a distro packager so I'm not the right person to ask really. But for each package the distros tend to have a file that looks something like this https://src.fedoraproject.org/rpms/python-tomli/blob/rawhide/f/python-tomli.spec where they describe build requirements, how to build, how to install etc. These files may need to adapt if build backend changes. Also for packages like build, flit_core, tomli, also hatchling are special cased a lot because they need to be built before the "regular" build tools exist.

Is it just a matter of adding the ability to e.g. apt install python-hatchling?

Yes this is another thing that needs to happen.

Don't quote me on any of this. I am NOT a distro packager 😄

ofek · 2022-01-06T16:50:52Z

Thanks! I'll work on that next.

ofek force-pushed the mypyc branch 5 times, most recently from 455414a to 8f8bef9 Compare January 5, 2022 05:08

Compile code with Mypyc

0ae8bca

ofek force-pushed the mypyc branch from 8f8bef9 to 0ae8bca Compare January 5, 2022 05:12

ofek commented Jan 5, 2022

View reviewed changes

disable Mypyc build hook by default

d810d5a

ofek force-pushed the mypyc branch from a8df365 to d810d5a Compare January 6, 2022 07:56

ofek closed this Jan 6, 2022

ofek deleted the mypyc branch January 6, 2022 16:51

ofek mentioned this pull request Jan 15, 2022

Compile code with Mypyc hukkin/tomli-w#33

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Compile code with Mypyc #165

Compile code with Mypyc #165

ofek commented Jan 5, 2022 •

edited

ofek Jan 5, 2022

ofek Jan 5, 2022

ofek Jan 5, 2022 •

edited

ofek Jan 5, 2022

ofek Jan 5, 2022

ofek Jan 5, 2022

hukkin commented Jan 5, 2022

ofek commented Jan 6, 2022

hukkin commented Jan 6, 2022

ofek commented Jan 6, 2022

hukkin commented Jan 6, 2022

ofek commented Jan 6, 2022

hukkin commented Jan 6, 2022

ofek commented Jan 6, 2022

		cont = cont[key_stem]
		return flag in cont["flags"] or flag in cont["recursive_flags"]

		relative_path_cont_keys = (header + key[:i] for i in range(1, len(key)))
		for cont_key in relative_path_cont_keys:

		# Pretend this exception was created here.
		TOMLDecodeError.__module__ = __name__

Compile code with Mypyc #165

Compile code with Mypyc #165

Conversation

ofek commented Jan 5, 2022 • edited

ofek Jan 5, 2022

Choose a reason for hiding this comment

ofek Jan 5, 2022

Choose a reason for hiding this comment

ofek Jan 5, 2022 • edited

Choose a reason for hiding this comment

ofek Jan 5, 2022

Choose a reason for hiding this comment

ofek Jan 5, 2022

Choose a reason for hiding this comment

ofek Jan 5, 2022

Choose a reason for hiding this comment

hukkin commented Jan 5, 2022

ofek commented Jan 6, 2022

hukkin commented Jan 6, 2022

ofek commented Jan 6, 2022

hukkin commented Jan 6, 2022

ofek commented Jan 6, 2022

hukkin commented Jan 6, 2022

ofek commented Jan 6, 2022

ofek commented Jan 5, 2022 •

edited

ofek Jan 5, 2022 •

edited