fix: remove raise that limits using --config with dicts

[lpla]

Description

Snakemake use through other Python scripts using its API class

>>> import snakemake
>>> snakemake.snakemake(snakefile='Snakefile', config={"this_option": "does_not_break", "test": {'this_dict':'will_break'}})

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/lpla/bitextorment/lib/python3.8/site-packages/snakemake/__init__.py", line 508, in snakemake
    config_args = unparse_config(config)
  File "/home/lpla/bitextorment/lib/python3.8/site-packages/snakemake/__init__.py", line 960, in unparse_config
    raise ValueError("config may only be a flat dict")
ValueError: config may only be a flat dict

allows setting any kind of parameters but dictionaries. The code this PR removes was triggering an exception when there is no issue in using dictionaries as a setting inside the Python dictionary given to the Snakemake class through config.

A simple test named test_dicts_in_config(), similar to the code above, is added to tests/testapi.py file to reproduce this issue. Test could be even simpler, as the Snakefile can be completely empty to reproduce the issue, but the one I wrote shows a rule using dicts through config working without the raise.

QC

• The PR contains a test case for the changes or the changes are already covered by an existing test case.
• The documentation (docs/) is updated to reflect the changes or this is not necessary (e.g. if the change does neither modify the language nor the behavior or functionalities of Snakemake).

[johanneskoester]

Thanks! Actually a test of the CLI is possible. This works via the shellcmd argument of the test run function, see this example: https://github.com/snakemake/snakemake/blob/main/tests/tests.py#L267

[lpla]

Hi. Sorry about the initial version of the pull request. I thought we were running Snakemake through a Bash script, but it was a Python one, so we were using the Snakemake API. I updated the description and added a simple test. Let me know if this style is correct, or if I could do anything else to fix it better or cleaner.

[lpla]

Also, I saw you edited the title as "feat:". Isn't this a bug?

[pierrebarbera]

Just want to second this request. I ran into this limitation yesterday; as it is right now I'd have to choose between changing the config values I want to alter to be flat, or something janky like having a template config file in which I do string replacement. Passing a (nested) python dict is so much cleaner. Hope to see it in the next version so I can use it in production!

[pierrebarbera]

Actually, there is a somewhat dirty hack to make this behaviour work correctly, by specifying some, possibly meaningless, dummy "config_args" value when calling snakemake from python:

snakemake.snakemake(
  ...
  # dict with just the values to be updated, properly nested as the config expects it
  config=config_overrrides,
  # some dummy config arg, as would normally be passed on the command line via --config-args
  config_args=["dummy=0"]
  )

This is because the throwing function unparse_config is only called if there is a config specified, but no config_args, see here

[sonarcloud]

Kudos, SonarCloud Quality Gate passed!   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

No Coverage information
0.0% Duplication