MRG: Speed up browse figure creation if scalings are provided

[hoechenberger]

This avoids loading data from disk or performing calculations if we don't actually need to calculate scalings parameters.

Also adds a scalings parameter to Report.add_raw().

Performance improvements:

this branch: 2.17 s ± 214 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)
main:        4.07 s ± 625 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)

with the following code:

# %%
from pathlib import Path
import mne

sample_dir = Path(mne.datasets.sample.data_path())
sample_fname = sample_dir / 'MEG' / 'sample' / 'sample_audvis_raw.fif'

raw = mne.io.read_raw_fif(sample_fname, verbose=False)
# raw.pick([0])
raw.pick_channels(['MEG 0111', 'MEG 0112', 'EEG 001', 'ECG 001', 'STI 014'])
report = mne.Report()

# %%
%timeit report.add_raw(raw=raw, title='foo', psd=False, replace=True)

x-ref #10107

[agramfort]

Lgtm. Thx 🙏 @hoechenberger

[larsoner]

This logic seems wrong now as the not was omitted, should probably still be

Suggested change

	if isinstance(value, str) and value != 'auto':
	if not (isinstance(value, str) and value == 'auto'):

[larsoner]

... if you were trying to distribute the not, you probably meant:

Suggested change

	if isinstance(value, str) and value != 'auto':
	if not isinstance(value, str) or value != 'auto':

[hoechenberger]

Actually I believe we could probably change this this to

        if value != 'auto' and not isinstance(value, float):

WDYT?

[hoechenberger]

Pushed a respective change in cbc14e8

[larsoner]

I like the old conditional better, for example:

>>> import numpy as np
>>> np.array([1.]).item()
1.0
>>> float(np.array([1.]))
1.0
>>> np.array([1.]) == 'auto'
<stdin>:1: FutureWarning: elementwise comparison failed; returning scalar instead, but in the future will perform elementwise comparison
False

That warning would not happen with the isinstance(thing, str) call version of the code

To me the logic of if not (string and equal to 'auto'): then convert to float is clearer and more general

[hoechenberger]

@larsoner Please merge if you're happy!

[larsoner]

Otherwise LGTM

[larsoner]

You shouldn't need this line. _handle_default should EDIT: already make a copy internally

[larsoner]

(at least if you're worried about the user passing a dict and it being modified, it shouldn't be -- the values in principle could be if they're mutable and modified inplace but in practice this shouldn't really happen I think)

[hoechenberger]

I've removed the deepcopy now.

[larsoner]

Thanks @hoechenberger !