google / hrx Goto Github PK

Hi! I've recently been working on a Rust implementation for this, and found a few corner cases where the syntax from the README didn't match up with the examples:

The straightforwardest one is probably the duplicate-despite-quotes.hrx part of example/invalid/duplicates.hrx – it looks like (syntax says naught, #1) quotes are no longer specialcased?

The next one I ran into was duplicate-files.hrx from that same file – that archive should, according to the spec, (a) be valid and (b) contain <======> file\n.
I think so due to the following: contents is defined as "any sequence of characters that does not include U+000A LINE FEED followed immediately by boundary", and file as boundary " "+ path newline body?.
Now, given a buffer containing

<======> file
A      BCD  EF
<======> file

We can see, that the AB span matches boundary, C – the spaces, DE – path, and F – newline. What is left? To match the optional body, which consists of the following:

<======> file

Note, how this chunk doesn't start with U+000A LINE FEED, despite the line starting with boundary. This means, that the file contents continue until EOF.

The third mismatched example plagues example/empty-file.hrx. Assuming the same symbols as before, we get (after the first comment)

<===> file1
A   BCD   EF
<===>
So is this one.
<===> file2

thereby hitting the first LF+boundary sequence on the line declaring file2 (my parser returns {file1: {cmt: "This file is empty.", ctnt: "<===>\nSo is this one."}, file2: { cmt: null, ctnt: "" }}, which I feel is correct, going solely by the syntax?).

My hunch as to why these weren't noticed earlier is due to the usage of splitting parsers (e.g. in hrx.js and hrx.py), which probably handle these examples as expected.

I'd be more than happy to submit a PR addressing these issues, if deemed valid :)

I recently had the idea of creating a plain text archive format, searched the web and found this project, which is precisely what I was looking for. I needed a Java implementation though and decided I would implement an I/O library for Java for hrx instead of rolling my own custom archive format. I just found the time to do it yesterday and here are the results: https://github.com/topobyte/hrx-java

I implemented a test case for each of the example files, however I'm not 100% sure my implementation behaves exactly as specified / the same as the ruby implementation concerning trailing newlines. I think it would be very useful to have a set of testing files as a reference for current and future implementing developers. As a start, how about creating an extracted version of all the example files, i.e. we could have a directory extracted-examples containing a directory X for each X.hrx from examples. I'm not too fluent in setting up ruby and gems, so I would appreciate if you could create such a reference data set by unpacking all examples into corresponding directories and pushing it here. Thanks!

What do you think about having a list of implementations somewhere in the README?

I have found those so far:

• Ruby: https://github.com/google/hrx-ruby
• Python: https://github.com/rebeccajae/hrx.py
• JavaScript: https://github.com/rebeccajae/hrx.js
• Rust: https://github.com/nabijaczleweli/hrx.rs
• Java: https://github.com/topobyte/hrx-java

Are there any more that you're aware of?