Improvements to IO::Buffer read/write/pread/pwrite.
#7826
Conversation
ioquatix
force-pushed
the
io-buffer-write
branch
4 times, most recently
from
b6ebbbe
to
11b415d
Compare
|
I want to write some more tests. |
ioquatix
force-pushed
the
io-buffer-write
branch
2 times, most recently
from
d6354ec
to
243f1b5
Compare
| @@ -2508,23 2617,12 @@ rb_io_buffer_read(VALUE self, VALUE io, size_t length, size_t offset) | |||
| static VALUE | |||
There was a problem hiding this comment.
I think documentation needs update (at least length bytes, length is optional).
There was a problem hiding this comment.
I've updated the documentation for read/write/pread/pwrite and it should be more consistent: #7860
| @@ -2629,41 2727,55 @@ rb_io_buffer_pread(VALUE self, VALUE io, rb_off_t from, size_t length, size_t of | |||
| static VALUE | |||
| ssize_t result = read(argument->descriptor, argument->base, argument->size); | ||
|
|
||
| if (result < 0) { | ||
| return rb_fiber_scheduler_io_result(result, errno); |
There was a problem hiding this comment.
rb_fiber_scheduler_io_result returns either the read bytes or errno but not both, which can be problematic if a read fails in the second or later iteration. But as a workaround the caller can call read with length=0 and manually handle the retries if it needs this info, so I think this is a good trade-off, because the alternative would be to return both the read bytes and the errno, which would make an ugly api (errors would no longer be exceptions).
There was a problem hiding this comment.
I agree with you and I'm not sure if there is a better way to return "partial read or partial write error". I think if you ask to read or write at least N bytes, if you don't get N bytes, it's an error. Otherwise, as you say, you should just manage the read/write process yourself.
As discussed in socketry/async#224.