Serialization of the Parser Tree

[Goutte]

✨ Feature request

Make the parser tree serializable with the \serialize() function.

Motivation

Making the parser tree takes most of the total computation time. If I could "cache" it somehow, it would be great !

Example

$tree = $parser->program();
print(\serialize($tree));

yields

Serialization of 'Antlr\Antlr4\Runtime\Comparison\Equivalence@anonymous' is not allowed

Alternatives

• Make the grammar more efficient (there's room, I'm a noob)
• Use a visitor to transpile to target language (but code gen is always tricky)
• Make my own AST from the parser tree, and somehow ensure it is serializable (contexts?)
• How do YOU do it ?

Additional context

The vt time is the visitor. The rest could be cached, perhaps. On more complex scripts, with nested flow statements, it goes up to seconds.

Cyprak is the name of our budding DSL for an online role playing game (and is a pun on sprak == language)

[Goutte]

I managed to (un)serialize the parse tree, after some tweaking.

There are two issues:

• The anonymous class (easily solved by naming it)
• Trouble when unserializing because a single parser file (the generated parser) declares two namespaces (and one of them is not PSR-4). This can be hotfixed using unserialize-callback-func, but something could perhaps also be done upstream.

I can make a MR for the first part, if you're open.

[Goutte]

I went ahead and pushed the changes, so I can use my own fork.

Caching is worth it, this script takes about 60ms without it. You can see the ca (cache) and the vt (visitor) now eating up all the time