On 5/24/2016 12:16 PM, Anon wrote:
As if winning the compactness war while still being a symbol linkers won't have problems with wasn't enough, this approach also avoids generating giant symbols in the first place. Compression and/or hashing cannot do that. If D really wants to, it can compress symbols, but it *needs* to fix this problem properly *first*.
Because identifiers are generated from substrings that don't know about each other, to do your algorithm a scan of the combined string will still be necessary.
In any case, you should be able to implement your algorithm as a replacement for id_compress(). How about giving it a go?