On Saturday, 2 March 2013 at 10:05:08 UTC, Daniel Murphy wrote:
"Zach the Mystic" <reachbutminusthisz...@googlymail.com> wrote
in message
news:bueceuemxqmflixkq...@forum.dlang.org...
On Saturday, 2 March 2013 at 04:28:40 UTC, Zach the Mystic
wrote:
You would definitely need an identifier translation table:
"Dsymbols *" -> "Dsymbol[]"
Might as well just define ArrayBase etc
"NULL" -> "null"
Sure, but what about all the places 0 is used to mean NULL?
`//printf("...%d...", s)` -> `writef("...%s...", s)`
Why not just keep it as printf?
"#ifdef XIFDEFVERSION" + nested ifdefs + "#endif"
-> "version(XIFDEFVERSION) {" + nested {}'s + "}"
"#ifdef 0" -> "version(none)"
No luck, dmd source uses #ifdefs mid-declaration,
mid-statement, and
mid-expression (even mid-string-literal)
It also uses #ifs with complex conditions.
And don't forget no-args ctors, implicit calling of ctors,
stack allocated
classes, new keywords, narrowing integer conversions,
'virtual', pure
virtual function syntax, macros as expression aliases, string
literal
constness, the EXP_CANT_INTERPRET cast hack, macros,
namespaces, macros,
structs using inheritance, and of course more macros.
Every single one of these would have to be special-cased. If you
had a domain-specific language you could keep track of whether
you were mid-declaration, mid-statement, or mid-string-literal.
Half the stuff you special-case could probably be applied to
other C++ projects as well.
If this works, the benefits are just enormous. In fact, I would
actually like to "waste" my time trying to make this work, but
I'm going to need to ask a lot of questions because my current
programming skills are nowhere near the average level of posters
at this forum.
I would like a c++ lexer (with whitespace) to start with. Then a
discussion of parsers and emitters. Then a ton of questions just
on learning github and other basics.
I would also like the sanction of some of the more experienced
people here, saying it's at least worth a go, even if other
strategies are simultaneously pursued.
