Searching in a long line can be slow

[gwsw]

Searching for text near the end of a very long line can be unreasonably slow. For example, run

seq 30000 | tr '\n' ' ' | less

This produces a file containing one line which is 168,894 chars long. Type /29999 to search for a string near the end of the line. On a fairly fast desktop system this takes 160 seconds to complete.

[gwsw]

Fixed in 236e041.
The seq test now takes less than 1 second instead of 160 seconds.

[gwsw]

Note: 236e041 was buggy when used with non-ASCII text.
Really fixed in 3f8d4ab.

[mohd-akram]

This is still slow for me, but it was never 160 seconds. Performance is the same between versions (on a MBP 2016):

less 643:

$ less --version | head -n1
less 643 (POSIX regular expressions)
$ seq 30000 | tr '\n' ' ' | time less +/29999 +q
       10.80 real         7.49 user         0.12 sys

less 651x:

$ ./less --version | head -n1
less 651x (POSIX regular expressions)
$ seq 30000 | tr '\n' ' ' | time ./less +/29999 +q
        9.25 real         6.21 user         0.07 sys