Move to a real markdown parser

toast commented

2022-07-08 22:21:38 +00:00

Owner

There is https://akkoma.dev/sfr/marked-mfm, though it may need changes or forking for use.
Another possibility is forking https://github.com/commonmark/commonmark.js or writing a moo+nearley parser, but this should only be considered if neither of the above work (commonmark is not a particularly easy to parse language).

There is https://akkoma.dev/sfr/marked-mfm, though it may need changes or forking for use. Another possibility is forking https://github.com/commonmark/commonmark.js or writing a moo+nearley parser, but this should only be considered if neither of the above work (commonmark is not a particularly easy to parse language).

toast added this to the (deleted) project 2022-07-08 22:21:38 +00:00

Johann150 commented

2022-07-17 19:35:21 +00:00

Owner

I think this would be a chance to rework recognition of mentions which is currently a bit annoying because of recognizing mentions where it should not. For example in shell prompts, email addresses and matrix handles.

I think we could use this JavaScript RegExp to find mentions in a piece of text:

/(?<=^|\s)@([-.\w]+(?:@\w+(?:\.\w+)*)?)(?=\s|$)/gu

Perhaps it would also make sense to refactor the API for creating notes a bit: Replace the noExtractMentions parameter with something where you can pass in a list of mentions (and maybe their text representation separately like in ActivityStreams tag by convention?). If we want to do that we could also pass in a list of mentions into the MFM parser to make it understand if something should be recognized as a mention or not.

Also checking if a mention extracted with the regex are accounts that actually exist.

I think this would be a chance to rework recognition of mentions which is currently a bit annoying because of recognizing mentions where it should not. For example in shell prompts, email addresses and matrix handles. I think we could use this JavaScript `RegExp` to find mentions in a piece of text: ```javascript /(?<=^|\s)@([-.\w]+(?:@\w+(?:\.\w+)*)?)(?=\s|$)/gu ``` Perhaps it would also make sense to refactor the API for creating notes a bit: Replace the `noExtractMentions` parameter with something where you can pass in a list of mentions (and maybe their text representation separately like in ActivityStreams `tag` by convention?). If we want to do that we could also pass in a list of mentions into the MFM parser to make it understand if something should be recognized as a mention or not. Also checking if a mention extracted with the regex are accounts that actually exist.