bpo-40521: Per-interpreter interned strings #20085
Conversation
|
This change requires https://bugs.python.org/issue39465 to be fixed, and also to make the type method cache compatible with subinterpreters. |
With this PR, tests using subinterpreters like test_ast.test_subinterpreter() does crash since the method cache is built by default, whereas it shares interned strings between multiple interpreters. MCACHE must not be defined to test this PR, or the method cache must be fixed for subinterpreters. |
|
I rebased my PR on master. It is still a draft: PR #20058 must be merged first, and then the method cache must be fixed for subinterpreter, before we can consider to merge this PR. |
vstinner
changed the title
|
And I pushed a second fix for test_repl:
It's more an enhancement than a fix. Previously, the code also failed. It's just that previously, we didn't check if interned is NULL at exit, whereas my PR adds such assertion. |
Make the Unicode dictionary of interned strings compatible with subinterpreters. Remove the INTERN_NAME_STRINGS macro in typeobject.c: names are always now interned (even if EXPERIMENTAL_ISOLATED_SUBINTERPRETERS macro is defined). _PyUnicode_ClearInterned() now uses PyDict_Next() to no longer allocate memory, to ensure that the interned dictionary is cleared.
Make the Unicode dictionary of interned strings compatible with subinterpreters. Remove the INTERN_NAME_STRINGS macro in typeobject.c: names are always now interned (even if EXPERIMENTAL_ISOLATED_SUBINTERPRETERS macro is defined). _PyUnicode_ClearInterned() now uses PyDict_Next() to no longer allocate memory, to ensure that the interned dictionary is cleared.
…GH-20085)" (GH-30422) (GH-30425) This reverts commit ea25180. Keep "assert(interned == NULL);" in _PyUnicode_Fini(), but only for the main interpreter. Keep _PyUnicode_ClearInterned() changes avoiding the creation of a temporary Python list object. Leave the PyInterpreterState structure unchanged to keep the ABI backward compatibility with Python 3.10.0: rename the "interned" member to "unused_interned". (cherry picked from commit 35d6540)
…". This reverts commit 35d6540.
Make the Unicode dictionary of interned strings per-interpreter to
make it compatible with subinterpreters.
https://bugs.python.org/issue40521