* We weren't properly passing `z_tab` through to the underlying OBJECT
handler, which was causing PhpRedis to crash if you tried to execute
the OBJECT command in a pipeline.
* Rework the `testTouch` unit test to try and avoid erroneous failures
due to CI instance CPU scheduling.
Instead of just checking whether or not something is listening on the
high ports, send a quick PING to make sure.
We're not just using another Redis instance because the point of the
test is to protect against a regression when connecting to high port
numbers.
Add a mechanism that allows users to specify an arbitrary class name, in
combination with a search path so that PhpRedis unit tests can be run
against a different client.
Additionally, this commit allows multiple classes to be invoked in one
test execution either by passing multiple `--class` arguments, or a
class argument with a comma separated value.
* Use our `redis_cmd_append_sstr_key_*` and `redis_cmd_append_sstr_zval`
wrappers, which handle key prefixing and serialization transparently.
* Rework ZADD so it can handle the bulk double response from the `INCR`
options.
* Implement `ZDIFF`, `ZINTER`, `ZUNION`, `ZMSCORE`, and
`ZRANDMEMBER` for `RedisCluster`.
* Refactor `ZUNIONSTORE` command and switch to using our centralized
zset option parsing handler.
See #1894
* Create inline wrappers of the low-level php_stream_* functions that
also keep track of the number of bytes written/read.
* Change the logic to aggregate network traffic until the user
explicitly "resets" it. I think this will be a more common use-case
(running many commands and then seeing overall network IO).
See #2106
- Finish adding docblocks with examples for all of the stream commands.
- Fix XAUTOCLAIM response handler (the reply has a slightly different
structure to XCLAIM.
Implement the TOUCH command and refactor several of our "variadic key"
commands, which were previously all using their own specific handlers.
While refactoring the code, I changed `EXISTS` to require one key (it
had previously been set to require zero keys).
Additonally, it looks like we had a disparity in two commands which
should be idential to PhpRedis: SINTERSTORE and SUNIONSTORE.
Previously, SINTERSTORE required only one argument but SUNIONSTORE 2.
I simply changed SUNIONSTORE to also only require a single argument,
since that argument could be an array.
```php
$redis->sInterStore(['dst', 'src1', 'src2']);
$redis->sUnionStore(['dst', 'src1', 'src2']);
```
Refactor XGROUP and implement the new DELCONSUMER (Redis 6.2.0) and
ENTRIESREAD (Redis 7.0.0) options. Additionally, add a proper phpdoc
block to the stub file.
See #2068
* Add ZRANGESTORE command.
* Add Redis 6.2's `REV`, `BYLEX`, and `BYSCORE` to ZRANGE options.
* Refactor several ZRANGE family commands into a single reply and
options handler, using PHP's new argument parsing macros.
* Extend our tests to use the new ZRANGE options.
See #1894
Refactor the slowlog command to use the new argument parsing API and
also change it so we no longer manually handle atomic/non-atomic
logic in the handler itself.
Redis 7.0.0 allows for getting and setting multiple config values as an
atomic operation. In order to support this while maintaining backward
compatibility, the CONFIG command is reworked to also accept an array of
values or keys and values.
See: #2068
Update BITPOS to use the new argument parsing macros as well as
implement Redis 7.0.0's BIT/BYTE modifier.
Additionally I changed `int $bit` to `bool $bit` as its a more
appropriate datatype.
See #2068
With this commit, I can now run the full `--class Redis` unit test suite
inside of a PHP8.2.x buiild tree and have them pass and not throw any
fatal zpp argument errors.
Implement the new Redis 7.0.0 commands to pop multiple elements from one
or more lists/zsets.
Additionally, remove INTERNAL_FUNCTION_PARAMETERS from the
redis_sock_read_multibulk_reply_zval helper function as it wasn't
actually being used.
These changes allow the PHP 8 unit tests to pass even when zpp strict
mode is enabled.
I'm not exactly sure which setting causes the issue, but without these
changes I get many `zpp` errors if I run the tests inside of a PHP build
tree.
Instead of testing that `OBJECT IDLETIME` returns exactly zero after
updating a key in our tests, just make sure PhpRedis returns us some
numeric value.
We can't really test for a specific number of elapsed seconds in CI as
the VMs may be randomly preempted at times beyond our control, causing
the tests to "fail" even though there was no actual failure.
