Differential D6424 Diff 23458 swh/perfecthash/build.py

Changeset View

Standalone View

swh/perfecthash/build.py

This file was added.

				from cffi import FFI
				ardumontUnsubmitted Done Inline Actions same remark about headers. ardumont: same remark about headers.

				ffibuilder = FFI()

				# cdef() expects a single string declaring the C types, functions and
				# globals needed to use the shared object. It must be in valid C syntax.
				ffibuilder.cdef(
				"""
				typedef struct shard_t shard_t;

				shard_t* shard_init(const char* path);
				int shard_destroy(shard_t* shard);

				int shard_create(shard_t* shard);
				int shard_object_write(shard_t* shard, const char* key, const char* object, size_t object_size);
				int shard_save(shard_t* shard);

				int shard_load(shard_t* shard);
				int shard_lookup(shard_t* shard, const char* key, char** object, size_t* object_size);

				"""
				)

				ardumontUnsubmitted Done Inline Actions Isn't this ^ redundant with the header file `hash.h`? If not, i gather that's some of those tiny details about using cffi stuff that escapes me ;) ardumont: Isn't this ^ redundant with the header file `hash.h`? If not, i gather that's some of those…
				dacharyAuthorUnsubmitted Done Inline Actions It is not redundant and that confused me too at first :-) This is parsed by cffi and it does python binding based on it. Only what needs to be exposed is copy/pasted here. There would be a way to avoid the duplication but I'm not sure it's worth the effort: if there is an inconsistency it shows at compile time: no risk of divergence. dachary: It is not redundant and that confused me too at first :-) This is parsed by cffi and it does…
				ardumontUnsubmitted Done Inline Actions Then, the right effort needed is just to update a bit the existing comment, one extra sentence like this would do: The following is only the necessary part parsed by cffi to generate python bindings. What do you think? ardumont: Then, the right effort needed is just to update a bit the existing comment, one extra sentence…
				ffibuilder.set_source(
				"_hash_cffi",
				"""
				#include "swh/perfecthash/hash.h"
				""",
				sources=["swh/perfecthash/hash.c"],
				douarddaUnsubmitted Done Inline Actions is there some funky indent business here? douardda: is there some funky indent business here?
				dacharyAuthorUnsubmitted Done Inline Actions Yeah, fixed! dachary: Yeah, fixed!
				include_dirs=["."],
				libraries=["cmph"],
				) # library name, for the linker

				if __name__ == "__main__":
				ffibuilder.compile(verbose=True)
				olasdUnsubmitted Done Inline Actions Shouldn't this be prefixed by `swh.perfecthash` (and the module name be `swh.perfecthash._hash_cffi`)? olasd: Shouldn't this be prefixed by `swh.perfecthash` (and the module name be `swh.perfecthash.
				dacharyAuthorUnsubmitted Done Inline Actions I'm unclear about the requirements of ffibuilder: it works as it is therefore I conclude it should not be different. But I can try with prefixing the module with swh.prefecthash if you think it is important. dachary: I'm unclear about the requirements of ffibuilder: it works as it is therefore I conclude it…
				olasdUnsubmitted Done Inline Actions I mean, this looks like this is creating a "non-namespaced" extension module called `_hash_cffi` (which would be installed at the root of the python modules installation path), which sounds wrong. olasd: I mean, this looks like this is creating a "non-namespaced" extension module called…
				dacharyAuthorUnsubmitted Done Inline Actions Good catch: it is exactly like that and I overlooked it. I'm surprised it did not just break with tox. Fixed! dachary: Good catch: it is exactly like that and I overlooked it. I'm surprised it did not just break…
				vlorentzUnsubmitted Not Done Inline Actions nitpick: shouldn't it be called `swh.perfecthash._hash`? that's how the stdlib does it (eg. https://archive.softwareheritage.org/swh:1:cnt:92a71e14c480bdcab1fd43ecbfcd70bb6e23f0d5;origin=https://github.com/python/cpython;visit=swh:1:snp:3326130e6e26bc726591d44c04ec4d8d77cb20e8;anchor=swh:1:rev:fa26245a1c1aa938cce391348d6bd879da357522;path=/Lib/random.py;lines=58 ) vlorentz: nitpick: shouldn't it be called `swh.perfecthash._hash`? that's how the stdlib does it (eg.
				dacharyAuthorUnsubmitted Done Inline Actions Interesting. Since you're asking my preference would be to keep the name as it is but that's because I'm lazy. I'd be happy to change it if you think it's better though: I'm not that lazy ;-) dachary: Interesting. Since you're asking my preference would be to keep the name as it is but that's…
				vlorentzUnsubmitted Not Done Inline Actions heh, it doesn't matter vlorentz: heh, it doesn't matter