Closed NebularNerd closed 4 months ago
Not sure why checks are failing, still a bit of a GitHub noob when it comes to certain aspects.
EDIT: Fixed it, ran the code block I modified through Black Playground and all is well. Will have to look more into that later for debugging/prettifying my own code.
Conflicts resolved 🙂
Closes #57
While digging through files for PR #58 I discovered that we could do with a
footer
style reverse lookup on the multi part match to help boost confidence scores (especially if the footer is small as well). This modifies the multiple match to allow reverse looksup.To behave like the normal forward match, we aggregate the
matched.byte_match
andmagic_row.byte_match
to provide a longer match to the_confidence
function. One downside of the is that in the match data it will report the two matches smooshed together, e.g.:.mlt
is0x3c6d6c74
/<mlt
(0.4) and3c2f6d6c743e0a
/</mlt>\n
(0,7), combined confidence of 0.8 but you'll see<mlt</mlt>\n
in the data which looks untidy (could there be a better way to address this)Also, some unexpected results:
.pt2
cannot score above 0.8 when combined, I imagine it's because both are long matches which score 0.8 individually.ct
gets 0.4 forCREM
and 0.8 forDONE
but the combined will not go above 0.8Sample files.zip
Any improvements/comments are welcome, it works but there might be a better, nicer looking way to handle this.
New entries for magic_data.json:
.mlt
opensource video editor.ct
: Tracker format from the creator of BeRo Tracker BeRo's Blog, under Audio software.pt2
: Another music format from the creator of BeRo Tracker Picatune 2