[Scilab-Dev] Object Orientation

[Clément David]

Hi Amanda and Samuel,

Le mardi 21 février 2017 à 22:34 +0100, Samuel Gougeon a écrit :
> > But i don't understand how sci_macr2tree.c separates the function code from all SciLAB's memory.
> > I think the SciLAB macro memory it's like a binary stream without any markers that define where
> > a macro starts or ends.

Just to clarify and provide more information about the global design of Scilab 6 by my
understanding.

There is only one module called "ast" [1] involved in parsing, interpreting, values (or types)
definition. Sub-directories (both inside includes/ or src/cpp/) implement theses behaviors.

A bison/flex parser is used to allocate Exp (for expressions) [2] based on the syntax, all the Exp
objects then represent an abstract syntax tree of any Scilab code.

To interpret the code, we do use intermediate bytecode but simply implemented a "visitor pattern"
that traverses the Exp objects interpreting each Exp and passing values around. So the old-school
switch/case bytecode interpreter is replaced by a `visit(Exp*)` function dispatch for each Exp
class.

Values allocated by matrix definitions (by the parser) or function calls are all defined by Type
objects [3]. And most of our gateways are now using these internal Type management rather than API
Scilab to distinguish internal API and external API (aka API Scilab).

Note: the macr2tree is implemented as a visitor that encode each Exp to a specific Type, it just
traverses the Exp tree to reproduce a Scilab 5 compatible values encoding.


[1]: modules/ast
[2]: modules/ast/includes/exps and modules/ast/includes/exps/exp.hxx
[3]: modules/ast/includes/types and modules/ast/includes/types/internal.hxx

Regards,

--
Clément