On 2020/04/27 22:57:06, dak wrote: > Move keytable init out of constructor. Frankly, I don't consider this an > improvement.
As anecdotal data, it took me about as long to get this working as the whole other patch, and I had it segfault about half a dozen times and had another half-dozen compiler errors while getting stuff right. And doc_hash_table, all_ifaces and option_hash are initialised in the same manner as what I did in the previous patch version. And it's not like lexer startup time is really going to factor in here a lot... The previous code generated an individual key table for every single lexer (except for cloned lexers). So without further feedback, I strongly lean towards pushing the previous version. https://codereview.appspot.com/577840053/
