Uri Guttman wrote: >>>>>> "SB" == Steve Bertrand <st...@ibctech.ca> writes: >
> i told you dispatch tables are very useful! > > i leave how to code it as an exercise for you. Well, I did end up implementing the dispatch table, and had some serious problems trying to retrofit it into a particular class's namespace, but it's done. Quickly I found that I was copy/pasting the way I did it, and will have two spots to make changes, but was able to do it without making any API changes. Even though I continued on to completion for experience, I realized that what I should be doing is configuring the 'struct' definitions themselves as the dispatch table (which I will proceed to do). My type definitions are a hash, where the key is the 'tag', and the value is a human-presentable description of the tag. I figure that if I make the 'description' contain a coderef to the validation directly, I've saved much time, and held true to the primary programmer virtue, all the while not needing to make any API changes. I have a semi-related code question that has to do with the tightening of param validation. In my ISP::Vars class, I have a compare() method, which now checks for extra, _and_ missing 'tags'. I've compiled in a config var that I can use while building the gui that states: [ISP::Vars] forgive_missing_attrs = 1 I want to add the ability for a 't/*.t' test to either disable this check, or implement some form of "IM_A_TEST" variable into the ISP::Var namespace without having to modify the config file itself. I'm sure this is a symbol table thing, but as always, loose direction that will lead me to further reading is welcome ;) Steve
smime.p7s
Description: S/MIME Cryptographic Signature
