The pec programming language FAQ

Q: How mature is pec?

A: It's not :) Good ideas/help is appreciated in all areas, from
language design, implementation, test cases, etc.  That being said,
you should be able to use the pec compiler to write prototype
applications (see examples in the test case directory).
Comments/questions/bug reports can be sent to brettletner at gmail dot
com.

Q:  Where did the name 'pec' come from?

A: A long time ago it stood for Pure Embedded Compiler.  pec is no
longer pure, but I still liked the name.  Now I guess it stands for
pec embedded compiler :)

Q: Why isn't pec pure?

A: Mostly I wanted to make it easier for C programmers to migrate.  A
pure language with the same target market seems like a fine thing,
though.

Q: How does the type system prevent out of bounds array indexing?

A: It uses Haskell classes and built-in "Count" types.  See the paper
"Lightweight static capabilities" by Kiselyov and Shan for more
information.

Q: Can I use the type system to prevent other common sources of
errors, e.g. divide by zero?

A: Yes.

Q: Why no garbage collection?

A: GC is generally (right or wrong) frowned upon in embedded
development.  It makes the run-time system more complicated and
makes integration with other GC'd languages more difficult.  A GC'd
language with the same target market seems like a fine thing, though.

Q: Why no recursion?

A: Recursion is generally (right or wrong) frowned upon in embedded
development.  Also, not having (general) recursion leaves open the
possibility of compiling the whole program into a state machine.

Q: Why no closures?

A: Pec doesn't have a heap, so it can't have "upward" closures.
"Downward" closures may be added in the future.  It does have a simple
form of macro lambda.

Q: Why no objects?

A: Personal preference.

Q: Why no parallelism/concurrency?

A: I just haven't thought about it yet.  My best guess is that it will
eventually be based on STM, but help is needed here.

Q: Why no operator precedence?

A: It is better style to just put the parens in and removes all code
review discussion as to whether you should have parens or not.  Also,
I was too lazy to implement it :)

Q: This "safe" pointer isn't safe at all, what gives?

foo = do
  p = new (Just 'a')
  case p of
    Nothing -> ()
    Just c -> do
      p <- Nothing
      putCh @c

A: The second assignment to p (i.e. p <- Nothing) should be disallowed
while c is in scope.  The compiler currently doesn't prevent this (and
other similar errors such as returning a pointer to the (soon to be
invalidated) stack).  If anyone has an idea on how to statically
prevent this I'd love to hear it.

Q: Why does the pec parser take so long to compile?

A: I don't know.  Help on this from a 'happy' expert would be much
appreciated.

Q: How do I modify the pec grammar?

A: You'll need my compiler construction tool, 'grm'.  Grm is similar
to bnfc in that it will generate a parser, pretty printer, and AST
from an input grammar file (e.g. pec.grm).  Grm can be found on
hackage and git@github.com:stevezhee/grm.git.

Q: At least one of the examples in the euler directory (Euler10.pec)
takes a very long time (infinity?) to compile under LLVM.  Why?

A: I don't know.  Help on this from a llvm/llc expert would be much
appreciated.