On 02/17/2017 01:42 AM, Glen Larsen wrote:

Interesting. Searching is pretty snappy, case insensitive, but doesn't ignore diacritical marks. The load time is comfortable for 627 entries and it would be interesting to see how load times would be for larger sets.

Maybe, if I can find the time, I might be able to set up a test case with the whole mutopia library as records to test larger numbers.

This is a client-side index so index build performance is also dependent on the client side processor, OS, and memory.

Yes, although it is possible to pre-compute the index(es) and download it(them) alongside the data itself. So a question of a trade-off between downloading time and processing time after download.

To do advanced search I think we might need an index for each field (so a composer index, an instrument index, etc.) to be able to query those fields individually. On the Clairnote site I just have one index with everything in it and everything is searched.

Here is an interesting anomaly: I searched on the year 1832 and got 2 hits, one of them being year of "Wolfgang Goethe (1749-1832)". When I search on the birth year (1749) I get 3 hits. As I wrote this and went back and forth with it I noticed that it is likely being affected by the unicode character "–" (u+0336) --- 1749-1832 versus 1749–1832--- but I'm not sure why.

I wonder how complicated it would be to normalize strings (ignore accents).

Good question. Maybe in the code that creates the index you could normalize the index entries as the index is created. Should be doable I think...?


Mutopia-discuss mailing list

Reply via email to