Unfortunately there appears to be exactly one place you can do this, and it
turns out to be inside inferTypes (because you need to know the type of c
completely, and you can't type-infer x until you know if it's a tag or a
variable). It's definitely nicer than doing it in the parser, though.
I've also started adding "-- AMBIGUITY" comments in the parser.
Type inference and checking is now handled entirely by the later passes.
There are a few remaining places in the parser that look at the types of things
that have been defined; this is in order to resolve syntax ambiguities (e.g.
c[x], c ! x). This is a temporary measure to minimise cgtest breakage; it can't
work for things that need to be inferred (e.g. "CHAN INT c:" "d IS c:" "d !
x"), so it'll need moving out to a pass in the near future.
There's still quite a bit of work to do on this, but results so far are
encouraging: the code is an awful lot cleaner, and about four hundred lines
shorter.
Previously the parser set it to the element type, but everything that used it
set it to the type of the whole array. Now it's documented in AST.hs, and the
parser makes it the type of the whole array, since that's almost always what
you really want later on.
It's redundant, since you can always compute them from the variable, and it
makes the code that deals with actuals rather cleaner.
On the other hand, it slightly complicates some of the tests, because any names
you use in an Actual need to be defined...
This touches an awful lot of code, but cgtest07/17 (arrays and retyping) pass.
This is useful because there are going to be places in the future where we'll
want to represent dimensions that are known at runtime but not at compile time
-- for example, mobile allocations, or dynamically-sized arrays. It simplifies
the code in a number of places.
However, we do now need to be careful that expressions containing variables do
not leak into the State, since they won't be affected by later passes.
Two caveats (marked as FIXMEs in the source):
- Retypes checking in the occam parser is disabled, since the plan is to move
it out to a pass anyway.
- There's some (now very obvious) duplication, particularly in the backend, of
bits of code that construct expressions for the total size of an array
(either in bytes or elements); this should be moved to a couple of helper
functions that everything can use.
This works at least for simple examples, although it's probably a bit
restrictive on the array indexes you're allowed; it should attempt to
constant-fold them.
This used to work by adding a magic prefix to the error message, but it appears
that doesn't work with the GHC 6.6 version of Parsec. It now searches for a
magic substring anywhere in the error message.
It uses // as a delimeter rather than \0 now, since including nulls in Strings
causes problems -- for example, putStr "a\0b" will only print "a".
This patch is actually an amalgam of multiple (already large) patches. Those patches conflicted (parameterised Structured vs. changes to usage checking and FlowGraph) and encountered a nasty bug in darcs 1 involving exponential time (see http://wiki.darcs.net/DarcsWiki/ConflictsFAQ for more details). Reasoning that half an hour (of 100% CPU use) was too long to apply patches, I opted to re-record the parameterised Structured changes as this new large patch. Here are the commit messages originally used for the patches (which, as mentioned, were already large patches):
A gigantic patch switching all the non-test modules over to using parameterised A.Structured
Changed the FlowGraph module again to handle any sort of Structured you want to pass to it (mainly for testing)
A further gigantic patch changing all the tests to work with the new parameterised Structured
Fixed a nasty bug involving functions being named incorrectly inside transformInputCase
Added a hand-written instance of Data for Structured that allows us to use ext1M properly
Fixed a few warnings in the code
ErrorReport is of type (Maybe Meta, String), thereby adding an optional code position to error messages.
Die has been changed so that die and dieP are now implemented in terms of dieReport (:: ErrorReport -> m a). This involved changing less code than changing die to be of type ErrorReport -> m a. All that had to be changed directly was that Die instances now implement dieReport instead of die.
Any bits of code that "caught" errors has been changed so that it handles ErrorReport instead of String. This ErrorReport is eventually, in Main, passed to dieIO, which will soon be changed to read the file in and provide the context. Accordingly, MonadIO m has been added as a constraint to dieIO, and dieInternal has been changed to no longer use dieIO (because really we can't add the MonadIO constraint to dieInternal).
Various error messages have been changed. Notably, all instances of fail in ParseOccam have been changed to use die or, wherever possible, dieP. A similar thing has been done in EvalConstants and EvalLiterals.
The function showCode shows code as either occam or Rain depending on the frontend. This is then used by a formatCode function that acts similar to
printf, which makes it easy to format error messages that use showCode.
