Description
Goal: to make it super lightweight to make a new test case so that you don’t think twice about it, while keeping them robust enough not to rot if we rework internals
-
New command
cursorless.recordTestCase, which asks for a test case name / maybe directory -
Waits for next cursorless command
-
On next cursorless command, takes a snapshot of current file, including cursor locations, visible lines, file type, and relevant decorations
-
Takes snapshot after cursorless command, including cursor and visible lines
-
Saves before, command, context, and after as a test case
-
Then add code that can parse this type of fixture and run it as a test
-
I'd argue that cursorless command should be captured pre-inference, but don't feel strongly
-
This technique can be used for testing transformations, inside / outside, actions, etc
-
For example, to check whether the delimiter / inside outside stuff is working for eg "type" transformation, can run the command with "take" and "Chuck". This makes it robust to changes in internal representations, at the expense of being slightly less granular
-
Also save highlights to make sure those don't break?
-
could then also use these test cases to automatically record GIFs (Make GIF generator using existing test cases #215)
-
let's use this one to capture Add type transformation #32 (comment) as a test case
Extra credit is:
- capture clipboard before and after
Follow-on issues filed:
- capture visible lines before and after Test a command's effect on visible ranges #160
- test ability to work across splits #216
- capture highlights Support testing decorations #199