autocomplete

---

Autocomplete v9

This is where all the changes to autocomplete are made. This readme will be updated for each version of autocomplete

Folders and Files

• fig/: Wrappers around the fig.js API to make it nicer to work with
• parser/: Parses command line output using a bash parser and then a contextual parser that incorporates completion specs
• generators/: Launches asynchronous requests to generator suggestions dynamically
• suggestions/: Utilities to compute, sort, and filter a list of suggestions from parser results and generators that have completed
• state/: Manages react state including figState from fig.js hooks, parser results, generator results, suggestions to display, and visibility of the window
• hooks/keypress.ts: Handles keypresses from the user, updating state accordingly

How it works

The primary logic for the app is contained within state/ and App.tsx which synthesize the parts listed above into the autocomplete app. In particular, state/ lays out a reducer that dictates all of the ways the state of the fig app can change. We want to update state at the following times:

• We receive an autocomplete event from the fig.js API. We call setFigState to update the current buffer, cwd, etc. in our autocompleteState
• The buffer has changed. We must re-parse the new buffer. This is done asynchronous because of the need to load spec files from disk. When the parser is complete we call updateParserResult.
• A generator has completed. If the results aren't stale, we must update our generator state with the new results, calling updateGeneratorResult
• We receive a keypress even from the fig.js API. Depending on the keypress, we might need to update the selected item index, or possibly hide the autocomplete window. We call hide or scroll to update state.
• An unrecoverable error occurs. We want to hide the fig window until we can show something again.

The functions/actions mentioned above are all defined in state/.

Every time we update the state in one of the ways mentioned above, we might need to get a new list of suggestions. We also might need to trigger generators if the search term that the user typed for the current argument has changed. Every time we update state, we do a few checks to see if we need to regenerate suggestions or trigger generators.

Versioning

New versions of autocomplete are created when changes are made to the autocomplete spec format. For example, if we add “generators” as a property of args, this is a change. AND changes to the fig hooks that are sent

Testing locally

You can launch the autocomplete engine locally by running:

npm run dev

which will start serving the repo with hot reloading on localhost:3000

Tip: to view logs in the console when their source is displaying instrument.ts, right click anywhere it says "instrument.ts" and click "Blackbox Script" in the option menu