romshark / icumsg / 38138386980

98%

push

github

fix: Find duplicate options in linear time

Check the name of each select and =n option against the names seen so
far instead of against all sibling options, which takes quadratic time
in the number of options. Search the first 8 names linearly and keep
any further names in a map, so that the few options of a typical
argument cost no more than before:

    {n, select, o0{x} o1{x} ... o39999{x} other{x}}  (380 KiB)

    Before: 3.7 s
    After:  6 ms

A plural with 40,000 =n options takes 5 ms instead of 3.7 s.
Add a select with 1000 options to BenchmarkTokenize.
It takes 0.08 ms instead of 2.6 ms,
and the other inputs show no difference.

When an argument repeats more than one name,
report the first option that repeats an earlier name:

    {n, select, a{1} b{1} b{1} a{1} other{1}}

    Before: ErrDuplicateOption at byte 27
    After:  ErrDuplicateOption at byte 22

29 of 29 new or added lines in 1 file covered. (100.0%)

1778 of 1820 relevant lines covered (97.69%)

911.76 hits per line