[issue21118] str.translate is absurdly slow in majority of use cases (takes up to 60x longer than similar functions)

[STINNER Victor]

STINNER Victor added the comment:

str.translate() currently allocates a buffer of UCS4 characters.

translate_writer.patch:
- modify _PyUnicode_TranslateCharmap() to use the _PyUnicodeWriter API
- drop optimizations for error handlers different than "ignore" because there is no unit tests for them, and str.translate() uses "ignore". It's safer to drop untested optimization.
- cleanup also the code: charmaptranslate_output() is now responsible to handle charmaptranslate_lookup() result (to decrement the reference coutner)

str.translate() may be a little bit faster when translating ASCII to ASCII for large string, but not so much.

bytes.translate() is much faster because it builds a C array of 256 items to fast table lookup, whereas str.translate() requires a Python dict lookup for each character, which is much slower.

codecs.charmap_build() (PyUnicode_BuildEncodingMap()) creates a C array ("a three-level trie") for fast lookup. It is used with codecs.charmap_encode() for 8-bit encodings. We may reuse it for simple cases, like translating ASCII to ASCII.

----------
keywords: +patch
Added file: http://bugs.python.org/file34691/translate_writer.patch

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue21118>
_______________________________________