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

[Serhiy Storchaka]

Serhiy Storchaka added the comment:

fast_translate.patch works only with ASCII input string and ASCII 1:1 mapping. Is this actually typical case?

Here is a patch which uses different approach. It caches values for ASCII keys. It works with all types of input strings and mappings and can speed up more use cases, including non-ASCII data, deletion and enlarging.

translate_timing.py results:

                                unpatched           patched
Testing 1-1 translation
str.translate                   4.55125927699919    0.7898181750006188
str.translate from bytes trans  1.8910855210015143  0.779950579000797
Testing deletion
str.translate                   4.481863372000589   0.7718261509999138
Testing enlarging translations
str.translate                   4.421521270000085   0.9290620680003485

translate_script_ascii.py results:

---------------------------+---------------------------+-------------------------------
Tests                      | translate_script_ascii.34 | translate_script_ascii.cached3
---------------------------+---------------------------+-------------------------------
replace none, length=10    |           6.12 us (+176%) |                    2.22 us (*)
replace none, length=10**3 |           448 us (+1293%) |                    32.2 us (*)
replace none, length=10**6 |           474 ms (+1435%) |                    30.9 ms (*)
replace 10%, length=10     |           5.73 us (+133%) |                    2.46 us (*)
replace 10%, length=10**3  |           412 us (+1060%) |                    35.5 us (*)
replace 10%, length=10**6  |           442 ms (+1204%) |                    33.9 ms (*)
replace 50%, length=10     |            4.75 us (+85%) |                    2.57 us (*)
replace 50%, length=10**3  |            311 us (+552%) |                    47.7 us (*)
replace 50%, length=10**6  |            331 ms (+617%) |                    46.2 ms (*)
replace 90%, length=10     |            3.36 us (+29%) |                    2.59 us (*)
replace 90%, length=10**3  |            178 us (+250%) |                    50.8 us (*)
replace 90%, length=10**6  |            192 ms (+291%) |                    49.2 ms (*)
replace all, length=10     |            2.64 us (+28%) |                    2.06 us (*)
replace all, length=10**3  |            146 us (+189%) |                    50.3 us (*)
replace all, length=10**6  |            152 ms (+194%) |                    51.7 ms (*)
---------------------------+---------------------------+-------------------------------
Total                      |          1.59 sec (+651%) |                     212 ms (*)
---------------------------+---------------------------+-------------------------------

----------
Added file: http://bugs.python.org/file34734/translate_cached.patch

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