• =?UTF-8?B?Ik5vcmRsw7Z3Ig==?= (10/10) Apr 08 2014 I'm looking for hot spots in the DMD source where deduction of
• =?UTF-8?B?Ik5vcmRsw7Z3Ig==?= (13/13) Apr 08 2014 So far I figured that what I typically want to do is
• =?UTF-8?B?Ik5vcmRsw7Z3Ig==?= (3/3) Apr 08 2014 I just realized that maybe the best way is to recurse down the

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

I'm looking for hot spots in the DMD source where deduction of 
types of expressions are performed.

I want to use this to extract information about the type of an 
expression before/after point.

Is this logic spread all over

.*Exp::semantic
.*Exp::semantic2
.*Exp::semantic3

or is there a hot spot where I can insert a callback to print 
such information?

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

So far I figured that what I typically want to do is

printf("Type: %s", e->type->toChars());

whenever I find a type-deduced expression e that fulfils

(e->loc->linnum == QUERY_LINE &&
  e->loc->charnum <= QUERY_COLUMN &&
  QUERY_COLUMN < (e->loc->charnum +
                  e->length_of_expression_in_code))

Can I somehow query the source code range (start end begin 
offset/row-column) of an expression?

This of course has to handle sub-expression types which is a 
complication I haven't covered here.

So maybe a better feature would be to query for type of 
expression before or after point.

=?UTF-8?B?Ik5vcmRsw7Z3Ig==?= <per.nordlow gmail.com> writes:

I just realized that maybe the best way is to recurse down the 
AST until I find node(s) whose token range cover my query point 
right?