* add @sourcebot/schemas package
* migrate things to use the schemas package
* Dockerfile support
* add secret table to schema
* Add concept of connection manager
* Rename Config->Connection
* Handle job failures
* Add join table between repo and connection
* nits
* create first version of crypto package
* add crypto package as deps to others
* forgot to add package changes
* add server action for adding and listing secrets, create test page for it
* add secrets page to nav menu
* add secret to config and support fetching it in backend
* reset secret form on successful submission
* add toast feedback for secrets form
* add instructions for adding encryption key to dev instructions
* add encryption key support in docker file
* add delete secret button
* fix nits from pr review
---------
Co-authored-by: bkellam <bshizzle1234@gmail.com>
* feat: move language utilites out of components and dedup linguist languages
This mot moves all linguist language and codemirror languages into
utility classes.
It also adds syntax highligting for 200+ languages which can use
available parsers.
* Add support for all available 3rd party languages
* Remove some accidental languages
---------
Co-authored-by: Brendan Kellam <bshizzle1234@gmail.com>
* feat: add all 800 linguist language icons
This uses existing svg icons for common languages, then falls back on
the iconify library for the remaining 600 languages.
* move icon component to ui components
---------
Co-authored-by: Brendan Kellam <bshizzle1234@gmail.com>
* feat: Add support for codemirror mode languages
Codemirror 5 had a number of languages supported via modes instead of
explicit language support
* Add additional languages, remove comment
The motivation for building search suggestions is two-fold: (1) to make the zoekt query language more approachable by presenting all available options to the user, and (2) make it easier for power-users to craft complex queries.
The meat-n-potatoes of this change are concentrated in searchBar.tsx and searchSuggestionBox.tsx. The suggestions box works by maintaining a state-machine of "modes". By default, the box is in the refine mode, where suggestions for different prefixes (e.g., repo:, lang:, etc.) are suggested to the user. When one of these prefixes is matched, the state-machine transitions to the corresponding mode (e.g., repository, language, etc.) and surfaces suggestions for that mode (if any).
The query is split up into parts by spaces " " (e.g., 'test repo:hello' -> ['test', 'repo:hello']). See splitQuery. The part that has the cursor over it is considered the active part. We evaluate which mode the state machine is in based on the active part. When a suggestion is clicked, we only modify the active part of the query.
Three modes are currently missing suggestion data: file (file names), revision (branch / tag names), and symbol (symbol names). In future PRs, we will need to introduce endpoints into the backend to allow the frontend to fetch this data and surface it as suggestions..
