DecompressionTest: improve tests

[jrfnl]

👉🏻 Note: the methods covered via this test class need a lot more tests, in particular the Requests::compatible_gzinflate() method, which includes code paths currently completely untested and handling of ZIP file format, which when called from Requests::decompress() (which is the only way it is called from within the Requests library itself), can never be reached as the "magic markers" used by ZIP are not included in the Requests::$magic_compression_headers property.

A future iteration should do a more in-depth review of these two method to determine what should be handled by each method and add additional tests. Still, this first iteration is already an improvement.

EncodingTest: move to subdirectory and rename

There are a number of test classes which all directly relate to the WpOrg\Requests\Requests class.

Having those test files/classes all in the tests root directory does not make it any clearer what these tests are actually testing. So, I'm proposing that if a class warrants multiple test classes, we create a directory named after the class and place the files in that directory.

Also, the tests in this file are not about encoded data, but about compressed data, so renaming the test class to be more descriptive.

DecompressionTest: reorder methods

Move the data providers (and associated helper methods) down to below the tests they apply to.

DecompressionTest: rename test method parameters

... to better reflect the data which is expected to be passed to the tests.

DecompressionTest: refactor data providers

• Rename the methods to use data as the prefix.
• Remove the static keyword as data providers do not need to be static (and haven't needed to be for a long time).
• Add keys to the data arrays to turn the data sets into named data providers.
• Add keys to the information in each data set to make the data providers easier to understand.
• Remove the "helper" methods to "create" the data providers. They are unnecessary and obfuscate what data was provided to the tests.
• Use all three data providers for each test instead of the helper method.

DecompressionTest: add extra test cases

... to ensure that non-compressed data is returned unchanged.

Requests: improve handling of non-compressed semi-empty strings

... in Requests::decompress() and Requests::compatible_gzinflate().

Bow out earlier for non-compressed strings containing only whitespace.

Includes extra test case and fixes the failing test for Requests::compatible_gzinflate() as introduced in the previous commit.

DecompressionTest: improve documentation

Includes adding @covers tags.

Requests::compatible_gzinflate: fix @since tag

The @since tag should reference the version in which the method was introduced in Requests, not in WordPress.

Related to #497

[schlessera]

I'm currently trying to figure out how we can cross-test these compressed strings.

The closest I've gotten so far:

echo 'foobar' | gzip -c | hexdump -v -e '"\\x" 1/1 "%02x"' ; echo

This produces the following output:

\x1f\x8b\x08\x00\x00\x00\x00\x00\x00\x03\x4b\xcb\xcf\x4f\x4a\x2c\xe2\x02\x00\x47\x97\x2c\xb2\x07\x00\x00\x00

It looks like the output formatting is correct, but the shell gzip produces a different result. Not sure yet whether this is due to the input, or the compression algorithm...

[jrfnl]

There are a number of places in the tests where we will need to do a cross-check that the test data is correct.
IIRC, I already did so for the IdnaEncoderTest and documented the tooling I used inline in the test, but there are more (like this one).

Maybe we should open a separate issue about this ?

[jrfnl]

Also: a number of these test cases were only recently added via #309 by @soulseekah . Maybe they can shed some light on how the test cases were created ?

[soulseekah]

These are the headers:

"\x1f\x8b" => true, // Gzip marker.
"\x78\x01" => true, // Zlib marker - level 1.
"\x78\x5e" => true, // Zlib marker - level 2 to 5.
"\x78\x9c" => true, // Zlib marker - level 6.
"\x78\xda" => true, // Zlib marker - level 7 to 9.

I don't remember how exactly I generated the output, it was on a Linux machine a very long time ago but probably with plain old gzip.

[soulseekah]

Also make sure you do echo -n I think that's the difference in your compression test, @schlessera - the newline character. This means more data and a different CRC (last bytes).

echo -n 'foobar' | gzip -c | hexdump -v -e '"\\x" 1/1 "%02x"' ; echo
\x1f\x8b\x08\x00\x00\x00\x00\x00\x00\x03\x4b\xcb\xcf\x4f\x4a\x2c\x02\x00\x95\x1f\xf6\x9e\x06\x00\x00\x00

Which matches the cases :)

[jrfnl]

Thanks @soulseekah for pitching in! That's helpful.

[schlessera]

Good catch, @soulseekah !

[schlessera]

@jrfnl The above command should be added as a comment to the tests so that we know how to generate new cases in the future.