robb / asterism Goto Github PK

This one is a little more tricky since NSCountedSet inherits from NSMutableSet.

Since we would want ASTMap(countedSet, block); to be equivalent to ASTMap((NSSet *)countedSet, block);, "unrolling" versions of map et. al. would need an extra parameter.

E.g.:

// if unroll is YES, calls the block multiple times per element.
NSCountedSet *ASTMap(NSCountedSet *set, BOOL unroll, id(^block)(id obj));

like the underscore.js equivalent

Not sure what should happen if the assumption is violated, tho.

Worth it?

Guess it's not worth the trouble after all and it's easy to bring back if need be

That may allow better integration with cocoadocs and Dash. Even though I personally prefer reading TomDoc in the headers, this may be the right thing to do for the community at large.

yay or nay?

https://twitter.com/jpstuehler/status/450956880819806208

Seems like there's little use for it once ASTLift is out, too. (See #23, #24)

… for NSArray and NSOrderedSet

I find it odd that All and Any methods return NO for empty sets. Formally, I believe, YES should be returned in such cases; according to: https://en.wikipedia.org/wiki/Vacuous_truth#Vacuous_truths_in_mathematics

Additionally, I find I often need to cover an empty-container case manually in my code, because Asterism doesn't return YES in those cases.

Tests probably shouldn't import the commands directly, to make sure this is caught in the future.

I'm having second thoughts about asserting against nil all over the place, I think it's best and more in the spirit of the surrounding Cocoa framework if stuff like

ASTMap(nil, ^(id obj, NSUInteger index) {
     return [obj foo];
});

simply returned nil.

Essentially:

• Explain people never to link against Asterism, because that will fail if clang ever changes the name mangling of overloaded C methods

or:

• Move everything over to static inline defined in the .h files

/cc @wrl

Consider methods that operate only on id<NSFastEnumeration> such as ASTAll

ASTAll(@{ @"key": @"value" }, someTest);

This will actually evaluate @"key" rather than @"value", which is probably more desired.
However, if we add

BOOL ASTAll(NSDictionary *dict, BOOL(^block)(id value));

later, this may lead to some nasty surprises. Better to pre-empt that now and add dictionary versions that call the existing ones with allValues.

Affects

• ASTAll
• ASTAny
• ASTGroupBy
• ASTIndexBy
• ASTMin
• ASTMax
• ASTPluck
• ASTReduce

✨

Write specs where necessary to make sure that ASTWhatever(nil, block) == nil.

yay or nay?

Those methods would make sense for NSIndexSet:

• ASTAll
• ASTAny
• ASTDifference
• ASTEach
• ASTEmpty
• ASTFilter
• ASTFind
• ASTHead
• ASTIntersection
• ASTMap
• ASTMin
• ASTMax
• ASTReduce
• ASTReject
• ASTSize
• ASTTail
• ASTUnion
• ASTWithout