Skip to main content

False Negative in Search

Completed
Jack Brannen

This is an odd one but I'm finding specific false negative on search.

Create this exact string as a single file in the current NV database:

“Popped corn”
 
Then search NV for "popped." My results do not return the new text file created above.
 
macOS 12.4
nvUltra Version 1.0.0 (2022.11.02.23.49)
0

Comments

9 comments

Date Votes
  • Fletcher Penney

    I assume you just had typos in your post, so perhaps you can clarify more precisely.  If you search for "popped." (in quotes, with a trailing period) in a file that does not contain "popped" followed by a period, it will not match.

    But again, I am presuming errors in your post rather than your search??  If so, you can always delete the index and force a reindexing to ensure that the index is up to date.

    As an aside, the way the files are indexed can result in a very rare rate of false positives, but not false negatives (assuming "proper" searches).

    0
  • Jack Brannen

    Hey Fletcher, you are correct—my """correct""" English punctuation made for a technically confusing post. My exact search term was

    popped

    How do I delete the index?

    0
  • Fletcher Penney

    You can delete all cached indices with the button in the General Preferences.  Then reopen the folder and it will be indexed again.

    0
  • Jack Brannen

    Thank you, I was having trouble finding it!

    So I did successfully cause it to reindex (I saw the progress bar) and I am seeing the same behavior.

    0
  • Fletcher Penney

    If you’ll send the actual file in question attached to a support ticket I will take a look.

    0
  • Jack Brannen

    Thanks! Here it is.

    https://www.dropbox.com/s/njyu4eha5gnt7be/HMM%20Wickham%20Way%20Paint.md?dl=0 

    0
  • Fletcher Penney

    Thanks.

     

    Ok, what was happening was that you were using non-ASCII quotes, so they were being treated as "letters" instead of "punctuation."  So what was being indexed was actually `<left double quote>popped`.

     

    I have rewritten the indexing algorithm using a Unicode library so that only recognized letters and specific punctuation (dashes, connecting punctuation such as underlines) are included in the words that are indexed.

     

    The new algorithm passes the prior test suite, as well as a new test for the situation you discovered.  With some very rough tests, performance *seems* to be comparable.  I'll test it out a bit further before pushing it out.

    0
  • Jack Brannen

    Thanks for the expert bug-squashing!

    0
  • Fletcher Penney

    Thanks for reporting this!  Surprised it didn't come up sooner...

    0

Please sign in to leave a comment.

Didn't find what you were looking for?

New post
Powered by Zendesk