On 25/03/2014 01:03, Marvin Humphrey wrote:
Here's an example:parcel name: org.apache.foo parcel nickname: foo class name: search.BodaciousFrobnosticator class nickname: BoFrob method name: Do_Stuff = FOO_BoFrob_Do_Stuff Without the use of nicknames, that symbol would have to be something like `ORG_APACHE_FOO_BodaciousFrobnosticator_Do_Stuff`. I came to regret using the term "cnick" for this purpose and I'm the one who put "nickname" into the .cfp parcel JSON. I'm +1 on purging "cnick" from the Clownfish header language. This is ultimately just another installment in the ongoing Clownfish namespacing implementation saga. It's hard to get this right. Instead of replacing the keyword `cnick` with `nickname`, maybe it's time to revisit the option of eliminating class nicknames altogether (while keeping parcel nicknames).
So would you prefer to have method calls like 'BodaciousFrobnosticator_Do_Stuff'? I came to like writing 'VA_Push' or 'StrIter_Next' for frequently used stuff.
I'd rather get rid of parcel nicknames. What's the use of a hierachical namespace if we have an additional flat namespace with all its inherent problems? The parcel prefix can usually be omitted if the short names macro is in effect, so long prefixes aren't much of a problem.
That said, I don't want to make any radical changes at the moment. I'd propose to open a JIRA ticket with some of the ideas we had so far and revisit the namespace issue if we have support for another host language or a couple of Clownfish users.
Nick
