It also seems to work well as I use it so many times per day/week.

'Nuff said

Silence hint:Performance with an explicit copy.

Add a license file, some nim portability fixes, and finally move from stdlib Table to adix/lptabz.LPTab for a portable multi-Table.

Back to simple output rather than seekability testing with users able to configure what happens to any of 3 reasonable choices.

Fix problems/bugs exposed by nim-lang/Nim#17752 where a combination of not raising in urite and the new ignore SIGPIPE behavior made hldiff just barrel onward with failed write after failed write which was mostly a problem because this made hldiff continue to consume from a git log -p or hg log -p "to completion" even after a rapid quit out of a pager at the end of the differ|hldiff|pager pipeline, and running the differ/log generator can be very slow.

Swap version of index to length map by pointers for a 1.10..1.11x speed-up.

Fix bug found by @sschwarzer and generally make logic match the Python algo more strictly.

De-export a few things. hldiff is really much more program than library (unlike edits).

Use system.lines iterator which behaves better if input has \r \n line endings.

Guard against extreme slowdowns from hugely overlong input lines.

Speed up quite a bit (~2.5x) and get working in gc:arc/gc:orc mode.

This removes the #head dependency.

This should work now. Let me know if it doesn't.