Fix cue render order for ass/ssa.

[peerless2012]

This should fix #2124

• Add layer in cue and output.
• Add layer support for ssa parser.

[tonihei]

@icbaker Could you take a look? Thanks!

[icbaker]

I think we should default to 0 here for layer, same as Cue

That probably means you need to either return null here, or take a default value as a parameter, or don't define this general purpose function and just in-line it for the layer case.

I'd suggest in-lining

[peerless2012]

parseInt is a general method, and I will set layer to 0 when the result is unset in line 333.

[icbaker]

parseInt may be trying to be a general method, but LENGTH_UNSET (-1) is not a generally safe 'invalid' value for all possible integer fields (i.e. there may be some integer fields where we expect to read -1 from the file). In fact, tbh it's not clear to me that -1 (or any negative number) is an invalid value for layer. The SSA "spec" I'm reading says:

Layer (any integer)

Which could reasonably be interpreted to include negative numbers.

So please either in-line this method (my preference) or return null when the parsing fails.

[peerless2012]

@icbaker hello, please make a review for the new commit, and hope this pr can merge to 1.6.0 final release.

[icbaker]

Please can you add some tests for this?

I suggest:

1. Some parsing tests in SsaParserTest (including negative & unparseable layer values) - assert that the right value ends up in Cue.zIndex
2. An end-to-end test in SubtitlePlaybackTest? This will cover both your parsing code and the re-ordering in the CueGroup constructor.

For (2) I suggest you craft an SSA test file that has cues overlapping in time with different layer values, to demonstrate why the re-ordering has to be done 'outside' SsaParser.