Multiple messages with the same old name

[Pierre-Sassoulas]

Description

Permit to have multiple messages with the same old name. There are optimisations done that make Pylint a little faster.

Type of Changes

	Type
✓	✨ New feature

Related Issue

This is it, after this PR the real review for #1164 can start.

[coveralls]

Coverage increased (+0.007%) to 90.03% when pulling 877a68e on Pierre-Sassoulas:multiple-messages-with-the-same-old-name into ac25a27 on PyCQA:master.

[Pierre-Sassoulas]

I made some performance review (This is why I removed the MessageId class). Regarding performance, I think this is now making things better but it's not very clear cut:

Benchmark used

Pylint analysing itself (~= 65s) :

Total time

time python3 -m cProfile -s tottime pylint/__main__.py pylint/ > result

Sum of the time for the function containing "message_"

cat result|grep message_|cut -c 14-18|paste -sd+ | bc  

Pylint analysing astroid (~= 1,25s) :

Not shown here because it's too fast.

Before

Total time

68,30s user 0,46s system 99% cpu 1:08,81 total

Sum of the time for the function containing "message_"

.453

Longest message function

ncalls  tottime  percall  cumtime  percall filename:lineno(function)
40359    0.138    0.000    0.369    0.000 message_handler_mix_in.py:175(is_message_enabled)
51443    0.122    0.000    0.140    0.000 message_definition_store.py:152(get_message_definitions)
40359    0.070    0.000    0.091    0.000 message_handler_mix_in.py:197(is_one_message_enabled)
40358    0.028    0.000    0.028    0.000 message_handler_mix_in.py:186(<listcomp>)
9800    0.013    0.000    0.041    0.000 message_handler_mix_in.py:153(_message_symbol)

After

Total time

65,46s user 0,38s system 99% cpu 1:05,92 total

Sum of the time for the function containing "message_"

.573

Longest message function

   ncalls  tottime  percall  cumtime  percall filename:lineno(function)
 40396    0.166    0.000    0.542    0.000 message_handler_mix_in.py:173(is_message_enabled)
    51484    0.130    0.000    0.187    0.000 message_id_store.py:108(get_active_msgids)
    51484    0.097    0.000    0.325    0.000 message_definition_store.py:52(get_message_definitions)
    40396    0.066    0.000    0.085    0.000 message_handler_mix_in.py:195(is_one_message_enabled)
    51477    0.040    0.000    0.040    0.000 message_definition_store.py:60(<listcomp>)

Conclusion

It seems that overall the analysis is faster, but the time sepcifically for the part we changed is a little longer. We could probably merge MessageIdStore and MessageDefinitionStore to have less function call (51484 call to get_active_msgids are added) but I think it would affect maintenability a little.

[Pierre-Sassoulas]

@PCManticore this one is ready for review, it should be a piece of cake after #2992 ;)

[Pierre-Sassoulas]

Hi, @PCManticore, I won't be super available from mid-September to the end of October, now would be the best time to review this and #3021 :)

[PCManticore]

Awesome stuff @Pierre-Sassoulas ! 👍Sorry for the delay in getting to this, I was out as well.