Weirdness and Bugs in file import of watched words

I’ve found some weirdness with the importing of auto-tagged and flagged words from a file.

  • When importing auto-tagged words wploading a csv file of 83 rows only imports 11 rows on the first go. I need to upload it twice to get all 83 items

  • Likewise, when uploading ‘badwords.csv’ as flagged words, I only get 16 of 364 rows imported on the first go. Re-uploading does import the whole list

  • Even weirder, when I first uploaded my auto-tag words and then uploaded ‘badwords.txt’ on the flagged words tab then part of the list of auto-tags got moved to the flagged word list

  • Also weird, if I do it in the reverse order (upload flagged words, then try and upload auto-tag words) then nothing is imported (the alert says items have been imported, but none show)

  • Related to all this, when repeatedly clicking to delete the now-incorrect flagged words I easily hit the rate limits. This produces this somewhat inscrutable error message (I know it’s a rate limit because refreshing the whole page produces a text error message mentioning the rate limit):


As context — the auto-tag feature is really neat, especially as it allows regex. It’s a shame it’s hard to manage to lists of words with the current UI because we have (potentially) quite long lists of things we’d like to auto-tag. Likewise for the flagged words – we’d like to use public profanity lists, but these are quite long too.

I think more clarity about what happens if you download and then re-upload csv files would help. If there was an option to ‘replace existing list with uploaded data’ then this would reduce the problem of managing updates.

Also, given that the auto-tag already supports regex to match inputs, it would very cool if you could also include matching groups in the output — i.e. do substitutions to select matching tags.

Our use case is that we have ~400 internal codes which all start with the same three letters, ABC, ABCD or XYZ and then 3 or 4 numbers. We’d like these codes to be tags to enable filtering content, so it would be neat to set up a auto-tag rule which said (ABC|ABCD|XYZ)(\d{3:4}) and then have it add tags which equal $1$2.

I am not sure if this is a bug or just some weirdness as you’ve already mentioned, but I’ve moved it to the bug category as it reads that way. What version of Discourse are you on?

Am on the latest version self hosted. I’m using safari 14.0 if that’s relevant to any front end code.