Implement v1 in terms of v2 #73
Conversation
|
This builds on top of #72. |
dsnet
force-pushed
the
v1-via-v2
branch
6 times, most recently
from
d02227f to
494e46b
Compare
| before, after = before[1:], after[1:] | ||
| } | ||
| } | ||
| t.Errorf("known failures mismatch (-old +new):\n%s", strings.Join(diff, "")) |
There was a problem hiding this comment.
why not use e.g. go-cmp? I understand as a dependency it may be a problem for integrating into std, but I assume std will only care about the finished product as a single commit, or perhaps a few commits to land the changes in chunks, and not the full history - so we can import other deps in the meantime I assume.
There was a problem hiding this comment.
Hmmm, I guess that's true. Given that I already wrote this diff, it's probably fine.
v1/stream.go
Outdated
| case '[': | ||
| return Delim('['), nil | ||
| case ']': | ||
| return Delim(']'), nil |
There was a problem hiding this comment.
can't these four use return Delim(kind), nil?
There was a problem hiding this comment.
Delim('[') is a constant, so the compiler can pre-allocate this.
Delim(k) isn't a constant, so there's likely going to be a heap allocation at return to box this in an interface.
There was a problem hiding this comment.
Ah, I keep forgetting this. Perhaps add a comment to that effect.
There was a problem hiding this comment.
Huh, interesting, https://go.dev/play/p/gIsvN2C0hmF doesn't allocate. It seems that for many releases, the Go compiler avoids allocating "small" integers in an interface as it's already pre-allocated values for them.
It seems that the entire range of possible values for a uint8 are pre-allocated.
There was a problem hiding this comment.
Yes! That was @josharian some time ago :) https://go-review.googlesource.com/c/go/+/216401
Remove all the core functionality of v1 and instead implement it in terms of v2 with the appropriate options specified. Many v1 test cases currently failure because the v1 emulation in v2 is not a sufficient reproduction of historial v1 behavior. The list of known failures are stored in v1/failing.txt, which allows such failing tests to be skipped. As the v1 emulation in v2 improves, the failing.txt list should eventually become empty.
Remove all the core functionality of v1
and instead implement it in terms of v2
with the appropriate options specified.
Many v1 test cases currently fail because the v1 emulation in v2
is not a sufficient reproduction of historial v1 behavior.
The list of known failures are stored in v1/failing.txt,
which allows such failing tests to be skipped.
As the v1 emulation in v2 improves, the failing.txt list
should eventually become empty.