Re: some questions re. Ada/GNAT from a C++/GCC user

[dewar@cs.nyu.edu (Robert Dewar)]

Glenn said:

>Most folks that responded to this missed something.  That it is simply
>easier to parse source code that has all of the declarations in one
>section while building a symbol table.  They're all there in a group,
>bundled together like in Pascal, making it much simpler to identify
>them and enter them in the symbol table.

That's simply a misconception. For correct programs, there is obviously
no difference in the two cases from a compiler's point of view.

For error recovery, the Pascal/Ada conventions are harder work for the
compiler, because it has to try hard to identify the proper breakpoint
between declarations and statements in mangled programs. If it gets
confused on this point, you can get a lot of junk cascaded errors.

If you have not written a compiler, you often find that your intuitions
as to what might or might not help a compiler implementation are quite
wrong. Even if you DO know a lot about compilers, you can find yourself
surprised. There were restrictions in Ada 83 that were supposed to make
life easier for the compier, but had exactly the opposite effect (e.g.,
for most compilers, the restrictoin on subunit names, now removed in
Ada 95).