State of Development?

[nerffan1]

Hello,

I briefly used this API and it worked nicely, but I didn't realize that it has not been updated in a long time. Is there any updates or comments on the state of this API and/or longevity?

[nbehrnd]

@nerffan1 For one I would agree, parts of the documentation are a bit old (e.g., Python2 print statements instead of of Python3 print functions). On the other hand, is there a specific functionality not working for you (in a separate virtual environment)? Last (but not least), note CIRpy, equally written by Matt, which equally works with Python 3.12. "Old" doesn't automatically equate to "defunct"l; issues reported still receive help.

[pcorpet]

@nbehrnd can you help to make progress on #53 to fix #52 please?

[nbehrnd]

@pcorpet Thanks to flake8 (version 7.1.1 in a virtual environment of Python 3.13.1) the error's line was identified. If you remove the additional .txt extension (used to pass as an attachment below), here is a momentary «patch» (plus an additional shebang) to address only this issue.

pubchempy_edit.py.txt

[pcorpet]

Thanks for picking up so quickly. I see how to fix it, however it's not very convenient when deploying in a prod environment. I would prefer to rely on a published version of the package.

[nbehrnd]

@pcorpet I agree with you and infer your preference is for a wheel, and if possible, an update on PyPI. To familiarize me a little bit further, I set up a new branch test in my local fork to add the edit of this morning, and to use the test bench set up by @mcs07 within a CI.

Some of the tests fail. I attribute this to the backlog i.e., how to address PubChem now, and changes in syntax on behalf of Python. Indeed, there are a couple of format issues Python 3.12 permitted with a warning, and Python 3.13 now considers an error (example). Nevertheless, it is possible to (re)create a Python wheel (here).

[pcorpet]

Yes, the main issue would be to update PyPi once it works.

[nbehrnd]

@pcorpet May I reach out for your insight and potential help?

In branch test of my fork, I started to work on the issues pytest recommends to edit. Departing from commit 9c718ab, my presumption was to convert the string function molecular_formula of the main script into a float in lines 821 onward, i.e.

    @property
    def molecular_formula(self):
        """Molecular formula."""
        result = _parse_prop({'label': 'Molecular Formula'}, self.record['props'])
        try:
            return float(result)
        except (ValueError, TypeError) as e:
            print(f"conversion of `molecular_weight` to float failed, {e}")
            return None

However, this fails and the out again / still is a string as checked running the edited script in a proper virtual environment. With slight variation of the MWE the landing page provides

$ python
Python 3.13.1 (main, Jan  3 2025, 10:26:34) [GCC 14.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from pubchempy_edit import get_compounds, Compound
>>> comp = Compound.from_cid(1423)
>>> print(comp.isomeric_smiles)
CCCCCCCNC1CCCC1CCCCCCC(=O)O
>>> 
>>> print(comp.molecular_weight)
311.5
>>> print(type(comp.molecular_weight))
<class 'str'>
>>> 

and I fail to recognize how to provide a reliable data type conversion here.

pubchempy_edit.py.txt
pubchempy.py.txt

[pcorpet]

Are you converting the molecular_formula or the molecular_weight?

[nbehrnd]

@pcorpet True, I erred for one check.

There was a lot of back-and-forth to moderately improve the outcome of the tests using GitHub's Ubuntu 24.04 runner to "tidy up" later arvo

default pytest
============================= test session starts ==============================
platform linux -- Python 3.13.2, pytest-8.3.5, pluggy-1.5.0
rootdir: /home/runner/work/PubChemPy/PubChemPy
configfile: pytest.ini
collected 63 items

tests/test_assay.py ....                                                 [  6%]
tests/test_compound.py ..................                                [ 34%]
tests/test_compound3d.py ...E..                                          [ 44%]
tests/test_download.py ..                                                [ 47%]
tests/test_errors.py FF.                                                 [ 52%]
tests/test_identifiers.py ..                                             [ 55%]
tests/test_pandas.py ..F....                                             [ 66%]
tests/test_properties.py ....                                            [ 73%]
tests/test_requests.py ....                                              [ 79%]
tests/test_search.py ..                                                  [ 82%]
tests/test_sources.py ..                                                 [ 85%]
tests/test_substance.py .........                                        [100%]
 
==================================== ERRORS ====================================

...


ERROR tests/test_compound3d.py::test_atoms_deprecated - pubchempy.PubChemHTTPError: 'PUGREST.ServerBusy'
============== 3 failed, 59 passed, 1 warning, 1 error in 38.09s ===============

Edit: The addition of a time.sleep(2) fixture between each test function in conftest.py reduces the load for PubChem, and improves the situation

default pytest
============================= test session starts ==============================
platform linux -- Python 3.13.2, pytest-8.3.5, pluggy-1.5.0
rootdir: /home/runner/work/PubChemPy/PubChemPy
configfile: pytest.ini
collected 63 items
tests/test_assay.py ....                                                 [  6%]
tests/test_compound.py ..................                                [ 34%]
tests/test_compound3d.py ......                                          [ 44%]
tests/test_download.py ..                                                [ 47%]
tests/test_errors.py ...                                                 [ 52%]
tests/test_identifiers.py ..                                             [ 55%]
tests/test_pandas.py ..F....                                             [ 66%]
tests/test_properties.py ....                                            [ 73%]
tests/test_requests.py ....                                              [ 79%]
tests/test_search.py ..                                                  [ 82%]
tests/test_sources.py ..                                                 [ 85%]
tests/test_substance.py .........                                        [100%]
=================================== FAILURES ===================================

...

============= 1 failed, 62 passed, 1 warning in 152.91s (0:02:32) ==============

[pcorpet]

Many new tests fail due to changes in the PubChem API.

[nbehrnd]

@pcorpet I experimented a little bit in my fork's dev branch; according to tox' preliminary results this should work for either Linux, Windows, Mac on one, and Python 3.10...3.13 on the other side (ref) if testing/accessing the database without a PubChem token proceeds extra slowly. On the other hand, Matt is back.

pubchempy_wheel.zip