| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
| |
... with -Wsign-compare.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Output the escaped content directly into the destination buffer,
avoiding the memory copy and associated overhead.
This finishes the series, and brings another 6.6-18.0% speed
improvement in test_perf, compared to the previous increase with using
a static exception map.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
|
| |
Instead of precomputing a buffer length, do only one pass, extending the
buffer as necessary.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
| |
This avoids various additions and subtractions on the fast path.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
|
|
| |
Most of the callers know how much the buffer needs extending, so don't
ask them for the total size: new _ul_buffer_reserve_size replaces
_ul_buffer_ensure_size, and the fast path is now inlined.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
| |
Use an integer instead of pointer, like in the other cases in buffer.c.
We won't need the return value in the future.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* Don't lose old msg on realloc() failure
* Don't dereference NULL on failure when
buffer = (...)
if (!buffer) { buffer->len = orig_len; ...}
* To avoid the above pattern, use integer return values:
"buffer" itself will not change in any case.
* Add missing error handling
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Make the map static, read-only and preinitialized, to avoid
thread-safety problems.
* Invert the map, making it non-zero for exceptional characters
(so that non-exceptional characters don't need to be mentioned)
* Tighten the set of exceptional characters, so that most of ASCII
goes through the fast path.
This increases the speed of test_perf_* by 7.7-20.6%.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
| |
JSON-c 0.9 does not parse \f, so fall back to \uNNNN for now.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
| |
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
|
| |
|
|
|
|
|
| |
Be a little more explicit instead of relying on the compiler to optimize
the variable away.
Signed-off-by: Miloslav Trmač <mitr@redhat.com>
|
| |
|
|
|
|
|
|
| |
Make the string escape use an ul_buffer_t, to avoid unnecessary memory
allocations. This greatly reduces the number of mallocs made, and thus
results in a noticable increase in speed.
Signed-off-by: Gergely Nagy <algernon@madhouse-project.org>
|
| |
|
|
|
|
|
|
|
|
| |
Instead of escaping a string, returning a new one and running strlen()
on that, use an output variable to store the length, since we can
easily calculate that during the escape process anyway.
This should shave off a tiny bit of CPU time.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
| |
When finalizing the JSON buffer, always NULL-terminate it at the
appropriate length.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
| |
With this patch, both JSON keys and values will be properly escaped.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
| |
|
|
|
|
|
| |
ul_buffer_finalize() was off by one, and instead of overriding the
trailing ",", it overwrote the trailing nul-byte.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
|
|
To make the library thinner, we now format JSON ourselves: there was
so little we used from json-c, that doing it ourselves is lighter and
faster.
With this commit, we have the architecture in place, with only a few
little things remaining to be done, such as escaping string values.
Signed-off-by: Gergely Nagy <algernon@balabit.hu>
|
