Tapioca RBI generation slow for parser gem

[jesse-shopify]

I noticed this a few days ago but have only recently dug into it. When generating an RBI for the parser gem the performance is much slower than any other gem I have come across.

time bundle exec tapioca gem parser

Output (tapioca 10.2, parser 3.1.3.0):

bundle exec tapioca gem parser 180.31s user 17.23s system 98% cpu 3:19.76 total

Output (tapioca 10.3, parser 3.1.3.0):

bundle exec tapioca gem parser 187.51s user 16.53s system 99% cpu 3:24.77 total

[KaanOzkan]

For gem RBIs we parse yard documentation by default https://github.com/Shopify/tapioca/blob/64c09e8bbc912e12916fd4adb719dde5f8109b68/lib/tapioca/cli.rb#L197-L200

Problem is parser gem generates huge ruby files when installed. Parsing of these files by YARD is quite slow.

LOC:

File                                                        blank        comment           code
-----------------------------------------------------------------------------------------------
./lib/parser/lexer.rb                                        1078           3252          20685
./lib/parser/ruby31.rb                                       1401            245           6712
./lib/parser/ruby32.rb                                       1396            242           6700
./lib/parser/ruby30.rb                                       1361            244           6491
./lib/parser/ruby28.rb                                       1346            244           6457
./lib/parser/ruby27.rb                                       1331            234           6349
./lib/parser/ruby24.rb                                       1112            196           5428
./lib/parser/ruby26.rb                                       1115            201           5405
./lib/parser/ruby23.rb                                       1107            176           5386
./lib/parser/ruby25.rb                                       1109            201           5384
./lib/parser/ruby22.rb                                       1105            176           5374
./lib/parser/ruby21.rb                                       1101            176           5334
./lib/parser/ruby20.rb                                       1096            177           5290
./lib/parser/macruby.rb                                      1024            168           4985
./lib/parser/rubymotion.rb                                   1016            167           4942
./lib/parser/ruby19.rb                                       1017            169           4940
./lib/parser/ruby18.rb                                        914            156           4638

I think the best solution here would be to allow configuration of doc option to be gem specific. We can then ignore parser gem from being parsed.

One alternative you can use right now would be to exclude parser, rbi and tapioca gems on your tapioca/config,yml. It's probably being pulled in by tapioca so you don't need its RBIs for typechecking since you don't use the gem in your code. Another alternative is turning off doc option for gem command.

[fcheung]

Jumping on here to say that a gem specific configuration of the doc option would be very useful for me too - current versions of the aws-quicksight-ruby generates huge amounts of documentation, which took tapioca around 45minutes to handle (and resulted in a 30Mb rbi file)

[KaanOzkan]

Closing this in favour of the more actionable issue https://github.com/Shopify/tapioca/issues/1361