Commits on Aug 11, 2021

1. Parallel static analysis by default …

   Modern PCs have multi-core. RuboCop has `--parallel` option, but it is
   not used by default. This means that there is extra core resources for
   static analysis because multi-core is not being used effectively.
   
   This PR will be parallel processing by default, so RuboCop will run
   faster in proportion to number of cores. Options that can be
   parallel processing, such as `--only`, `--except`, and others have the
   same default behavior.
   However, there are options that cannot be parallel processing, such as
   auto-correction and others, so they are serial processing as before.
   
   `--parallel` can be a hidden option for some users. This change is
   expected to reduce user's wait time for running RuboCop.
   
   Below is an example on my development PC.
   
   Before:
   
   ```console
   % cd path/to/rubocop
   ./exe/rubocop --display-time
   (snip)
   
   1293 files inspected, no offenses detected
   Finished in 61.09493000002112 seconds
   ```
   
   After:
   
   ```console
   % cd path/to/rubocop
   ./exe/rubocop --display-time
   (snip)
   
   1293 files inspected, no offenses detected
   Finished in 10.13813699997263 seconds
   ```
   
   If user's PC has a lot of cores, memory and RuboCop targets a lot of files
   to inspect, it will be effective.
   
   One concern is that parallel processing uses more memory. For example,
   Ruby's MJIT is off by default as an example of concerning about memory
   consumption in a low memory production runtime (e.g. Heroku's minimum Dyno).
   However, RuboCop is a development tool, I assume that it will be fine
   in most cases.
   
   If user want to use it in a low memory production runtime, user can
   specify `--no-parallel`.
   
   Note: JRuby and Windows are not supported based on rubocop#4537.

   

   koic committed Aug 11, 2021
   Copy the full SHA

   0739982 View commit details

   Browse the repository at this point in the history