Add new fuzzy quote matching implementation #2814
Conversation
Codecov Report
@@ Coverage Diff @@
## master #2814 +/- ##
==========================================
+ Coverage 97.75% 97.77% +0.01%
==========================================
Files 203 204 +1
Lines 7720 7764 +44
Branches 1708 1718 +10
==========================================
+ Hits 7547 7591 +44
Misses 173 173
Continue to review full report at Codecov.
|
| return 0.0; | ||
| } | ||
| const matches = search(text, str, str.length); | ||
| return 1 - matches[0].errors / str.length; |
There was a problem hiding this comment.
readability for me :)
| return 1 - matches[0].errors / str.length; | |
| 1 - (matches[0].errors / str.length); |
There was a problem hiding this comment.
I agree, although I did have to suppress Prettier's formatting.
| const matches = search(text, quote, maxErrors); | ||
|
|
||
| if (matches.length === 0) { | ||
| // All matches had more than `maxErrors` errors. |
There was a problem hiding this comment.
I could perhaps call this a "candidate" or "candidate match"
e.g.
// All candidates had more than maxErrors errors.
There was a problem hiding this comment.
Looking at this again the comment is pretty superfluous, so I just removed it.
Implement a `matchQuote` function which will be used to replace `dom-anchor-text-quote` for finding the best match for annotation quotes in the document text. The new implementation is based on the `approx-string-match` library and provides several improvements over the existing one: - Better performance when there are many differences between the quote and closest document text - It will be easier for us to tune the degree of mismatch allowed between the quote and document text and how candidate matches are ranked
robertknight
force-pushed
the
match-quote
branch
from
0920964
to
500e9bc
Compare
This PR is the first part of #2779. It adds the new fuzzy quote matching algorithm and tests. The next PR will then switch quote matching in the client to use the new implementation.
Implement a
matchQuotefunction which will be used to replacedom-anchor-text-quotefor finding the best match for annotation quotesin the document text.
The new implementation is based on the
approx-string-matchlibrary andprovides several improvements over the existing one:
Better performance when there are many differences between the quote
and closest document text
It will be easier for us to tune the degree of mismatch allowed
between the quote and document text and how candidate matches are
ranked