feat: enhance asm functions to support all* of the Fift-asm syntax
#855
Conversation
Note, that while it's possible to support the syntax via Ohm without constucting ASTs, it's impossible to support the AST generation without outsourcing that part of the parser to the recursive-descent one, which would also keep track of dictionary of words/instructions and the TVM stack, at least approximating it. [See this for more info](#837 (comment)).
novusnota
changed the title
asm functions to support all of the Fift-asm syntaxasm functions to support all* of the Fift-asm syntax
The important test with arbitrary words/instructions containing `}` in them was needed
Co-authored-by: Anton Trunov <anton@ton.org>
|
Changed the linked issue to #857. I'm willing to return to #837 later, preferably when we start moving to Tact 2.0 and a new parser for it. Alternatively, can tackle it after other remaining issues on 1.6.0 milestone for me to resolve (both here and for docs) P.S.: On the positive side of #837, our |
There was a problem hiding this comment.
We should definitely add more positive tests from some known Fift-projects.
Can we parse at least some part of Asm.fif?
Some examples can be taken from this list: https://docs.ton.org/develop/smart-contracts/examples#fift-smart-contracts and this list: https://github.com/ton-blockchain/ton/tree/master/crypto/smartcont.
And apply a small fix to the parser — we cannot prohibit shadowing or
removal of instructions because both are used in Fift library and
beyond.
However, that means that out of place `{` and `}` are possible when
shadowing them. See the comment in `embed-fift-fif.tact` about such
case, and why it's ok to move forward with it
|
@anton-trunov added Fift library to the test suite and made sure that we don't create unnecessary snapshots for its contents |
Note, that while it's possible to support the syntax via Ohm without constucting ASTs, it's impossible to support the AST generation without outsourcing that part of the parser to the recursive-descent one, which would also keep track of dictionary of words/instructions and the TVM stack, at least approximating it. See this for more info.
P.S.: all* in the title means that we support everything, but with two caveats:
}and{(when shadowing them, for example) will break the parsing. However, the non-shadowing uses, lists{ }and any words containing}inside are allowed as long as they are not just a single}.wordmay not parse well, as they might contain the}inside of them separated by whitespace, which means that it's another variant of the 1st caveat.P.P.S: Essentially, our parser only has one flaw, which stems from
{ }being used as a function block. If we were to use, say,"""or{{{, the problem would kind of mitigate itself. Just until someone defines the"""or{{{as a new word in their Fift program, and then we (and FunC), would inevitably fail to parse it. Until we have a stack-keeping parser, that is.Issue
Closes #857.
Checklist