improvements to teatest & golden

[lrstanley]

Some ideas that could potentially improve teatest/golden (and testing bubbletea apps in general) after I initially started using the 2 packages (teatest/golden):

• potentially exposing View(), so users can use TestModel for both golden and non-golden testing for alt-mode based apps. Right now it seems like 1 or the other.
  • Right now, this would require special logic in bubbletea to support, unless you watched the reader and captured every screen flush.
• TestModel having more WaitFor type logic, e.g., m.WaitForContains(tb, "foo", "bar", "baz") for testing that a program contains multiple outputs
• The -update test flag is a bit annoying to work with. e.g., can't recursively use it across all tests, because any tests that don't have the -update flag (e.g., regular go tests), it will cause those tests to fail because they don't define that flag. An alternative option should probably be supported. e.g., a UPDATE_GOLDEN=bool type env var, which will make things like UPDATE_GOLDEN=true go test -v ./... more seamless.
• Consecutive calls to the golden functionality in 1 test overwrites previous test values, which seems like a developer error and shouldn't be allowed. Alternatively, consecutive calls should add additional golden files, suffixed with a number to indicate a loop or multi-check-based golden file check. This is what libraries like https://github.com/gkampitakis/go-snaps do.
  • Alternatively, a simpler approach is supporting an optional name field, or an additional function where you can pass a name, to create multiple snapshots in a single test.
• It would be nice for golden to also support passing in a colorprofile, so it can strip/simplify colors, so changing of colors doesn't fail tests, if the user doesn't care about colors. e.g., passing in the ascii colorprofile. I have local code for this I can PR as well.

Right now, most users who use teatest that I've seen have a bunch of wrapper logic around it, so hoping to improve the experience a bit so it doesn't feel as annoying to test bubbletea UI logic

Also FWIW, https://github.com/gkampitakis/go-snaps does a lot of really cool stuff around snapshot-based tests that might be good inspiration. I would say we could wrap that library, but they have a bunch of dependencies, so probably wouldn't be worth it. Some of the things they do:

• First run of tests, when a golden file doesn't exist, it will create it. No need to tell it to do an update on the first run (also because an update can cause other things to update).
• Multiple calls to snapshots create multiple entries in the snapshot file.
• Diffs are ansi-aware and render colors.
• Supports custom folder names for where snapshots are stored.

[aymanbagabas]

These are great ideas! I would also like to add that Bubble Tea could use more integrated tests via x/vt and x/xpty where we can create headless virtual terminals and run Bubble Tea applications on.

Using VT would allow us to capture the terminal screen content at any given time, and send input events to applications to see how they react. Basically, simulate running applications on actual terminals using TTYs/Consoles.