lua-users home
lua-l archive

[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index] [Thread Index]



On 2015-08-01 12:22, Dibyendu Majumdar wrote:

> On 1 August 2015 at 08:23, Dirk Laurie <dirk.laurie@gmail.com> wrote:
> > 2015-08-01 1:05 GMT+02:00 Christian Thaeter <ct@pipapo.org>:
> >> Some silly idea, why not doing it the other way around and use
> >> existing lua compatible syntax for ravi.
> >>
> >>  local i__integer = 0
> >>  function (a__integer, b__integer)
> >>
> >> I know, we all hate hungarian notation, its overly verbose and
> >> failed in various ways how it is proposed/used in C/C++.
> >
> > Oh, the verbosity can be cured, by borrowing an idea from
> > Fortran. (To go with origin-1 indexing, the arithmetic "for", etc.)
> >
> > Give Ravi a function IMPLICIT (capitals, please! like C preprocessor
> > directives).
> >
> > IMPLICIT ("integer", "[^I-N]")
> >
> > If a name matches the pattern, values assigned to it must be
> > of the indicated type. Ravi would of course not treat IMPLICIT
> > as a function but as a compiler directive. It's a function to
> > allow Lua to do nothing.
> >
> > To make Ravi programs compile under Lua, one can do this:
> >
> 
> Hi,
> 
> Thank you for all the ideas. One thing to note about Ravi's optional
> static typing is that unlike other implementations such as Typed Lua
> or Typescript - once a value has a static type annotation, Ravi
> strictly enforces this - as otherwise the JIT compiler could not
> optimize the generated code. I need to use a solution that indicates
> the guarantees provided by the language.
> 
> Using naming conventions alone is problematic as someone may have used
> the naming convention already - and that would mean Ravi would
> misinterpret existing Lua code. I am trying to ensure that in
> interpreter mode at least, all valid Lua programs are valid Ravi
> programs (this does break to a point as Ravi has smaller values for
> some of the Lua limits due to the modified bytecode format).

Legacy code will not contain type annotations anyway, you can just
provide a commandline option to switch explicit type annotations on/off,
with possibly different implementations how to declare types (for
backwards compatibility its prolly best to be off by default, then only
implicit inferred types are used)

....
> The other suggestion to use a preprocessr to convert Ravi code to Lua
> code - so that standard Lua can execute it - that's fine, and I may do
> that - but it does require a user to explicitly invoke the
> preprocessor.

Thats a great idea, especially when one could translate his
personal preferred naming conventions to types annotations eventually.

	Christian