This PR is stale because it has been open for 30 days with no activity.

Thank you for the patch, and sorry that it was left without attention for so long. The sniffer was rewritten on the main branch (gh-83273) and this patch no longer applies.

It should be easier now: the sample is split into lines keeping their line breaks, so detection can count '\r\n', '\n' and '\r' in a _detect_lineterminator() method, next to _detect_doublequote(). Note that falling back to os.linesep makes the result platform dependent; '\r\n' looks better for a sample without any line break.

#155061 implements this on the new engine. The only detail not covered above: the terminator is chosen by a majority vote among the line endings, with a tie broken in the order '\r\n', '\n', '\r'.

Closing as superseded. Thank you again for the patch.