In addition to the simple search the project will start with, we also want an 'expert user' text based query builder that is driven by a parser and the autocomplete system provided by the middleware ( #4 ). While this will in practice be the UI for only the most advanced users, its a smaller work load to build and produces the same output, therefore we can start developing and refining the complex query systems sooner.
[ ] Re-entrant tokenizer. Tokens should supply their start and end position in the stream.
[ ] Re-entrant parser. Parse nodes should supply their start and end positions in the stream
[ ] Parser should fail gracefully and attempt to resume. Ignore as little of the users query as possible.
[ ] Autocomplete popovers.
[ ] Integrate with Middleware to supply autocomplete information, scoped where possible.
brehaut
commented
8 years ago
In addition to the simple search the project will start with, we also want an 'expert user' text based query builder that is driven by a parser and the autocomplete system provided by the middleware ( #4 ). While this will in practice be the UI for only the most advanced users, its a smaller work load to build and produces the same output, therefore we can start developing and refining the complex query systems sooner.
@brehaut recommends investigating a Top Down Operator Precedence based parsing scheme
https://brehaut.net/media/files/al/textbox-metrics/ may be a starting place for the autocomplete popover (its rough as guts though), and https://brehaut.net/media/files/search-proto/ may be a starting place for the parser and tokenizer. Note: Codemirror is just for the prototype, far too heavy weight for our needs.
Depends on #4, #6