unix coverage test failure is just a flaky test: 1 tests failed: thread/thread_gc1.py

Code size report:

   bare-arm:    +0 +0.000% 
minimal x86:    +0 +0.000% 
   unix x64:    +0 +0.000% standard
      stm32:    +0 +0.000% PYBV10
     mimxrt:    +0 +0.000% TEENSY40
        rp2:    +0 +0.000% RPI_PICO_W
       samd:    +0 +0.000% ADAFRUIT_ITSYBITSY_M4_EXPRESS
  qemu rv32:    +0 +0.000% VIRT_RV32

The downside here is that importing an extensible module is (a lot) slower than a non-extensible built in module. It's slower because it has to search the filesystem (everything in sys.path) first, and then fall back to searching built-in modules.

At Pybricks, we actually want this to be able to import umath for backwards compatibility purposes

Is that the only reason you need this change? Maybe there's another way to support this without slowing down the import.

Note that if this PR were accepted then the docs need updating to add these modules to the list of extensible ones; see the "Extending built-in libraries from Python" section in docs/library/index.rst.

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.44%. Comparing base (8f8f853) to head (f0ffe0e).
Report is 2 commits behind head on master.

@@           Coverage Diff           @@
##           master   #17670   +/-   ##
=======================================
  Coverage   98.44%   98.44%           
=======================================
  Files         171      171           
  Lines       22192    22192           
=======================================
  Hits        21847    21847           
  Misses        345      345           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

It's slower because it has to search the filesystem (everything in sys.path) first, and then fall back to searching built-in modules.

Ah, we didn't notice because we don't have a filesystem.

Is that the only reason you need this change? Maybe there's another way to support this without slowing down the import.

Sure, open to suggestions.

Maybe something like this (untested!):

--- a/py/objmodule.c
+++ b/py/objmodule.c
@@ -207,6 +207,12 @@ mp_obj_t mp_module_get_builtin(qstr module_name, bool extensible) {
         if (!elem) {
             return MP_OBJ_NULL;
         }
+        #if MICROPY_LEGACY_U_MODULE
+        elem = mp_map_lookup((mp_map_t *)&mp_builtin_module_map, MP_OBJ_NEW_QSTR(qstr_from_strn(module_name_str + 1, module_name_len - 1)), MP_MAP_LOOKUP);
+        if (!elem) {
+            return MP_OBJ_NULL;
+        }
+        #endif
     }
 
     #if MICROPY_MODULE_BUILTIN_INIT

Or can you simply do:

MP_REGISTER_MODULE(MP_QSTR_umath, mp_module_math);
MP_REGISTER_MODULE(MP_QSTR_ucmath, mp_module_cmath);

in your main.c?