raeez / decaf Goto Github PK

Candidates
* use a container (http://hackage.haskell.org/package/hashmap or http://hackage.haskell.org/package/unordered-containers)

Confirmed: Hash Maps are the current "industry standard" solution to fast lookups (as opposed to our current [SymbolRecord].

generalize accessing the symbol tables and add this code to SymbolTable.hs
Add a layer of indirection to the semantic checker monad's use of the symbol tables
Implement more efficient symbol tables (maps etc.)

Perhaps consider a more general way of rewriting the tree (by writing generalized 'walk-functions')

Also, flatten the AST a bit (it's way too verbose atm)
- collapse the many levels of hierarchy surrounding operators
- collapse expressions into a more abstract form

Implement http://hackage.haskell.org/packages/archive/base/latest/doc/html/Data-Traversable

We need to merge my work with Scott's work, but in order for this to go cleanly, we should probably establish:
- good coding standards (see http://www.haskell.org/haskellwiki/Programming_guidelines and http://www.haskell.org/pipermail/haskell-cafe/2010-September/084006.html)
- an understanding of how to structure complex projects / programs

additionally, the merge should theoretically take the best techniques from each of our respective works and combine them (establishing our ultimate coding style in the process)

as title

semi-illegal/illegal-43.dcf
semi-illegal/illegal-44.dcf

Generate a graph of the AST, and perhaps generalize graph rendering.

We currently map ALL '\t' -> ' ' characters in the input string to the scanner; this results in internal program strings having no tabs, when they should.

Possible Solutions:
Move eatFirst/eatNext in Decaf.Scanner to return 'unconsumed' input (i.e. move calculation of updated source position to a per-token-basis.