Re-examine assumption that we need bottom-up parsing

[AsturaPhoenix]

Current assertion:

It is less straightforward to get top-down parsing to handle rewrite rules since top-down parsers are guided by a map of productions to patterns. In the case of rewrite rules, these productions can potentially be expansions or dynamic, which would need to be matched against the pending production in a top-down parse. This is doable, but involves a potential inverse function/production analysis, or may turn into a hybrid top-down-bottom-up parse.

Actually, I think a rewrite can be performed top-down by backtracking to where the rewrite began. This may greatly simplify the parse algorithm if we can get it to reconsider all relevant decision branches when it does such a backtrack, and if we can get it to work with retroactive rules.

Also: keep reading http://dotat.at/tmp/gll.pdf

[AsturaPhoenix]

After thinking further, the original point remains valid. Being able to backtrack to handle a rewrite presupposes that the parser was able to recognize the rewrite rule at all, which in general is not the case without the aforementioned inverse function/production analysis, or equivalently manual language design that explicitly includes rewrite possibilities in top-down recognition. I don't think this is acceptable; it is still much easier to do the parse bottom-up and introduce stack-like semantics on top of that algorithm.