sof-tplgreader: don't use RE for pipeline filtering

[aiChaoSONG]

If we have two pipelines with id=1 and id=10, current
RE is not able to filter these two pipelines from each
other.

This patch deprecates RE and use string comparison for
pipeline filtering.

Signed-off-by: Chao Song chao.song@linux.intel.com

[plbossart]

SOFCI TEST

[plbossart]

Timeout of APL_NOCODEC is gone with retest.

I also tested this PR locally on UP2-nocodec and UpExtreme-nocodec, no issues reported.

No idea what the pylint error means or if there's a better way to write this in Python but the concept is sound. Thanks @aiChaoSONG for the fix.

[marc-hb]

I had to "reverse-engineer" way too much code to review this small fix :-( #714 is a pure doc fix to avoid anyone having to go through the same pain again. Among other things it warns about the missing boolean precedence.

if type(value) == list:

The previous code supported only type(value) == list so please don't introduce this new "feature". Certainly not silently when fixing something unrelated and... probably never.

# check if current pipeline is the one we want

check if current pipeline is AMONG the ones we want

[fredoh9]

Some of Pylint warnings are making sense.

tools/sof-tplgreader.py:3:0: W0611: Unused import subprocess (unused-import)
tools/sof-tplgreader.py:4:0: W0611: Unused import os (unused-import)
tools/sof-tplgreader.py:5:0: W0611: Unused import re (unused-import)

[marc-hb]

pylint warnings fixed in no functional change PR #714, please help review

[marc-hb]

The previous code supported only type(value) == list so please don't introduce this new "feature". Certainly not silently when fixing something unrelated and... probably never.

@plbossart told me he needs this fix urgently so I applied my own suggestion above and resubmitted #715 to supersede this. Apologies.

[marc-hb]

pylint warnings fixed in no functional change PR #714, please help review

[aiChaoSONG]

@marc-hb it is correct to exclude the non-list support, actually I spend some time to understand check = len ([em for em in value if re.match(em + '$|' + em + '[^a-zA-Z]', str(line[key]), re.I)]) > 0 too.

[marc-hb]

First, I'm sorry I rushed #715 before we could have this discussion, that's bad. I blame @plbossart, he made me do it! He's always pushing everyone to rush things out... :'-D

Second, you shouldn't have approved my doc fix #714 if you think the code was supposed to support both lists and non-lists! Because I'm loudly saying the opposite in doc fixes #714 I mixed that up with something else.

Now back to the main question. This function is invoked in only one other function: _filter_by_dict() that consumes the output of parse_dict(). In parse_dict(), value is produced by the expression value.strip.split(',')` which always produces a list even when there is a single element in the user input. I traced and verified that.

I did spent a little bit of time looking at the removed regex code and I don't understand how it would make sense if value was not a list but only a string. In that case em would iterate on every character of the input string, so for instance for { 'type' : 'capture' } then em would iterate like this: c, a, p, t,...... how would that make sense?