Re: Symbol tables and scopes

>>I can't lay any claim to originality, but I use a two-layer structure.
>>I have one "base" entry for a unque symbol (name). The "base" entry
>>contains a list of pointers to the actual symbol table entries, one for
>>each declaration, and the id of the block containing the declaration. I
>>look up the symbol and then scan the list to find the applicable
>>declaration for the current scope.


> How exactly do you find the block with the "applicable" definition, in
> an tree of scopes/blocks?
>
> It looks like inefficient to me, when for every currently visible scope
> the block id must be compared with the id's of *all* possible blocks?


I maintain a tree of blocks. The scope of a declaration is the block
it's declared in and all nested blocks that don't contain a
redeclaration. I walk backwards up the tree and, for each block, see
if the declaration list for this variable contains that block id. The
first one I hit is the applicable declaration. Probably not the most
efficient, and, as I said, probably not original. It has the
advantage that it works. I'm open to other suggestions.

Return to the comp.compilers page.
Search the comp.compilers archives again.