Support arbitrary markers

[cbrnr]

Fixes #103. This is a big (breaking) change, I decided to tackle this as follows:

• If the passed events is an np.ndarray, the previous conversion magic is still happening (i.e. type is set to "Stimulus" and integers are converted to "S 1" etc.).
• If the passed events is a list of dicts, markers are written as is.

This means that if someone needs to write "old-style" markers with custom types (such as "Response", "Comment", etc.), they must manually create a corresponding list of dicts.

I'm not sure if we need to change anything in MNE-Python, but I don't think so. It reads descriptions and types and combines them to "Type/Description", which I think should continue to work. The events from annotations heuristics will only work for "old-style" annotations, but that's OK I think.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (e1a6c79) 100.00% compared to head (bd0bb13) 100.00%.
Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #118   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            5         5           
  Lines          634       625    -9     
=========================================
- Hits           634       625    -9     

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

[sappelhoff]

Thanks Clemens, I will have time to review this later this week!

[cbrnr]

Here's something to consider: pybv can already export arbitrary markers as type Comment, so maybe there is no need for such a big change? Coming from MNE, the private _export_mne_raw() function already does the job (it sets the type of all markers to Comment).

[sappelhoff]

Stimulus, Response, and Comment are "standard" BrainVision types. I would not cut them out of the description here but rather relax the language from MUST to MAY or similar.

[sappelhoff]

may also be int, IMHO

[sappelhoff]

I think removing this would result in some lines that are not pretty, like:

Mk2=Comment,Stimulus/S253,487,0,0

when it could also (more sensibly, as currently done) be:

Mk2=Stimulus,S253,487,0,0

[sappelhoff]

Fine to remove this, but I would somewhere keep a documentation that "Stimulus", "Response", and "Comment" are the classic types in a BrainVision file that is written by BrainVision Recorder.

[sappelhoff]

Coming from MNE, the private _export_mne_raw() function already does the job (it sets the type of all markers to Comment).

I assume you are referring to this?

pybv/pybv/_export.py

Lines 126 to 130 in 79dceaa

	if etype in ["Stimulus", "Response"] and description.strip().isdigit():
	description = int(description.strip())
	else:
	# if cannot convert to int, we must use this as "Comment"
	etype = "Comment"

Here's something to consider: pybv can already export arbitrary markers as type Comment, so maybe there is no need for such a big change?

I personally am able to write all data that I want to write. I thought your argument in #103 convincing so I am fine if you really need this change. But if the changes here are not a big advantage for you, then we should just leave it as is IMHO.