search

lachholden / obsidian-recipe-view

View your Obsidian notes as interactive recipe cards while you cook.
https://obsidian-recipe-view.readthedocs.io/en/latest/
MIT License
88 stars 3 forks source link

Can't mix fractions and decimals in quantity #20

Closed benjamin-hodgson closed 4 months ago

benjamin-hodgson commented 6 months ago

Some of my recipes feature quantities which look something like this:

1.5 g (¼ tsp) vanilla paste

These quantities get scaled incorrectly. For example, when scaling by 3:

3 .15 g (¾ tsp) vanilla paste
image

Seems the 1.5 is getting tokenised as two quantities. As far as I can tell the bug appears to be here; deUnicodeFractions is returning 1 .5 g (1/4 tsp) vanilla paste. The regex splits the string into 1 and .5 g (1⁄4 tsp) vanilla paste.

I think .+ at the start of the second group is probably too permissive, since we don't expect to see inputs with anything other than numbers on either side of the fraction slash post-normalisation. I tried replacing it with \d+ (to read /(\d+)(\d+\u2044.+)\b/ig) and it seemed to give me good results for all the inputs I could think of, but would defer to you on the desired behaviour.

Example with the proposed fix (after scaling by 3):

4.5 g (¾ tsp) vanilla paste
image

Thanks!

lachholden commented 4 months ago

Thanks for the detailed issue (and also for the coffee :))

I'm looking into this further now – the purpose of that regex was to help with mixed numbers that use unicode e.g. "1½", which I think would be broken by your proposed solution. It definitely needs to be made more specific though, working on that atm!

lachholden commented 4 months ago

Turns out NFKD normalisation is already converting numbers fully to ASCII – my previous incorrect assumption was that it converted them to superscripts and so it was possible to still disambiguate e.g. 1 1/2 from 11/2 after normalisation. Now, instead, I've removed deUnicodeFractions entirely and will simply require callers to pre-normalise and for there to be a space in mixed numbers. Going to write some tests and check everything still works.

A possible future fix if I really desperately want mixed numbers without spaces to work (unlikely) would be to add a space around unicode fractions if it doesn't exist prior to unicode normalisation (but then there are a bunch more characters that need to be matched because the normalisation hasn't happened yet)