Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
- decode_scan: Don't have .Reset() re-play an erroring rune
- decode_Scan: Have RuneTypeEOF always be zero-width
- jsonparse: Don't replace syntax errors with RuneTypeEOF if there's no barrier
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
These shouldn't happen, but if they do, then let's make it easier to
debug.
|
|
|
|
|
|
This should have been done back in 8aa12d3 (struct.go: Cache
structIndexes, 2023-01-26)
|
|
|
|
|
|
|
|
|
|
|
|
The CPU profiler tells me that the encoder is spending a lot of time on
type switches.
|
|
We already have a wildcard, no need to invent a new state.
|
|
|
|
|
|
It was confusing.
|
|
It's only used with a rune.
|
|
|
|
|
|
All formatting changes are made by
./tools/bin/golangci-lint run --fix ./...
|
|
|
|
IDK if my eyes are worse, or something changed with my font stack, but I
can no longer visually distinguish “ and ” so use " and » instead.
Plus » psycologically registers as "kinda special", so it's less confusing
to see as an object-thing in a stack.
|
|
|
|
|
|
|
|
|
|
|