VCD parsers: memory/array words are declared as 'mem [0][31:0]' with a space, so a first-token parser merges all words into one name
In VCD files produced by FSDB-to-VCD converters (and some simulators), each memory word is declared with the index as a separate token, e.g. $var reg 32 <code> mem [0][31:0] $end. A parser that takes only the token after the id code as the name (mem) and treats the rest as the bit range collapses every word of the array into one ambiguous name. They then vanish from any by-name comparison, which looks like the dump has no memory words when it does. Fix: join every token between the id code and $end, then split off only the final [msb:lsb] range, so the name becomes mem[0]. Check: count the $var lines whose reference has more than one bracket group, and make sure the parsed name count includes them. Scalar and vector names stay byte-identical.