What's the state of this lib?

[azrafe7]

NOTE: A haxe user reached out to me via email to ask some questions about the lib, so I thought it would have been useful to report that info here.

Hello! Could you please tell me what is the state of hxPako library? Is it considered stable to use in production? What is left to be done there? When are you planning to move it to Haxelib repository?

Hi and thanks for exposing your concerns. Here's a summary:

• hxPako has been updated not too long ago to be in sync with pako 1.0.4 (latest rev as of writing). Notably, from pre-1.0.0 to 1.0.4, a subtle - but critical - bug has been fixed, and custom dicts support has been added.
• Tests from nodeca/pako have been ported too (you can run them with haxe testAll.hxml -<target specific args here>); bear in mind that hxPako doesn't directly support strings and arrays (as stated in the README Notes). They're all passing for different targets, but neko, python and php are quite slow when compared to js/cs/java/cpp, so some of the tests may timeout and be reported as failing (please run them on your machine to have a first-hand proof of this).
• A possible bottleneck - for those slow targets - is probably to be found in a couple of big switches/gotos/FSM that I had to rewrite using while(true)/break/continue to be supported across all targets - as haxe doesn't have gotos/break-labels. This has to be investigated.
• The lib should currently be on par with nodejs pako, in the sense that it shouldn't (again using the conditional here!) have more bugs than the original ver. Of course you should see for yourself if that holds true (and feedback would be GREATLY appreciated!).
• What's left to be done? Making use of this lib in more real-world apps and see how it behaves would be a first step to ensure correctness; and so is profiling it and find a way to speed things up (especially for the targets mentioned above).
• In all sincerity I'm not planning to move this on haxelib anytime soon... but that's because it would only be more time consuming for me to do proper haxelib releases. I'd advice to just assume that the latest github version is the one you should use ;P.

PS: I'm going to repaste all of this into an issue on github, in case someone else can find this info useful.

[azrafe7]

Additional notes:

• this lib - as a port of nodeca/pako - implements zlib (most of it) and aims to be binary compatible with output from that lib, while adding a simple/chunkable API on top of it
• if you want to handle .zip files (i.e. archives of multiple compressed files/folders) this may not be what you want (but see here, it could be of use); you can build it on top of this though...
• no encryption/password protection is supported
• compressing might be slow (highly depends on target: see timings), decompressing should be ~fast enough everywhere (YMMV, as always: test to be sure)
• future-wise this lib could be improved upon by using native implementations where possible (ensuring output is still binary compatible with zlib)

[lublak]

future-wise this lib could be improved upon by using native implementations where possible (ensuring output is still binary compatible with zlib)

currently untested solutions:

for java || php || neko || hl || cpp || flash

haxe.zip.Compress/haxei.ip.Uncompress (InflateImpl)

A pure Haxe implementation of the ZLIB Inflate algorithm which allows reading compressed data without any platform-specific support.

for python zlib

for nodejs zlib