hotfix: make timestamp in msg_prefix for metadata regex compile optional

[SiQube]

as raised in #939 some people might want to extract the timestamp of a custom metadata type. I currently don't know if this would be possible. this PR provides a patch which fixes the problem raised in #939 while maintaining all other properties of compile_patterns.

I've additionally added a test which reproduces the problem. the code change of interest is in src/pymovements/utils/parsing.py.

feel free to reject this proposed change with a strategy to pass the new test (test_extract_timestamp_for_metadata) in tests/unit/gaze/io/asc_test.py and comment the solution on #939. additionally, we should double check documentation s.t. people can easily extract custom (meta)data from ascii files.

resolves #939

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (87f0780) to head (8d3230c).
Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #940   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           74        74           
  Lines         3448      3448           
  Branches       621       621           
=========================================
  Hits          3448      3448           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dkrako]

@SiQube Thanks for taking care of #939 !

The metadata_patterns argument is intended to be used with patterns that are only expected to be parsed once per file. This issue is related to #907

I guess that this part of our code renders your workaround unusable for use cases where the messages have multiple occurrences witch changing values:

pymovements/src/pymovements/utils/parsing.py

Lines 340 to 351 in 8d3230c

	elif compiled_metadata_patterns:
	for pattern_dict in compiled_metadata_patterns.copy():
	if match := pattern_dict['pattern'].match(line):
	if 'value' in pattern_dict:
	metadata[pattern_dict['key']] = pattern_dict['value']
	else:
	metadata.update(match.groupdict())
	# each metadata pattern should only match once
	compiled_metadata_patterns.remove(pattern_dict)

If we take something like the message MSG 2055307 TRIALID TRIAL 0 MOVIESTART, your test pattern (?P<movie_start>\d+).*?(?P<movie_id>\d+)\s+MOVIESTART would match multiple times and reassign the values to the metadata field, losing the prior values in the process.

So I guess this is change is not harmful and can be still useful as a workaround to get the timestamp of a metadata pattern.

In case of multiple matches per file this would only contain the last matched values. I would rather propose an additional message_patterns argument to parse messages, or extend the patterns argument to unify message patterns, metadata patterns, and sample column patterns as written in #939 (comment)

[SiQube]

I closed this, let's work on #939 instead