Show HN: Underscore.js for Objective-C | Hacker News Reader

Show HN: Underscore.js for Objective-C

underscorem.org

120 pointsrobert-boehnke25 comments

This brings to mind google's warning re: the use of functional idioms in Java: http://code.google.com/p/guava-libraries/wiki/FunctionalExpl... Objective-C blocks make things smoother than Java, but it still feels to me like the language wants you to have a bunch of small objects sending message to each other. (rather than doing function composition)

That said, I use underscore.js in nearly all of my browser and node.js code, so I'll certainly consider this when the iPhone train comes back around.

If you embrace the Smalltalk roots of Objective-C, you'll be a lot happier, and "all those braces" are completely comprehensible.

No, you need to embrace how Objective-C is written today because blocks (and all those braces) are a fundamental part.

I think that's what I was trying to say, but apparently didn't make it clear.

There are also ports of Underscore.js to Perl & Lua:

* http://vti.github.com/underscore-perl/

* http://mirven.github.com/underscore.lua/

And AS3, if you're into that sort of thing (self-plug): https://github.com/amacdougall/underscore.as

Interesting if you're already used to underscore.js.

The more Objective-C way to do this, without the dot syntax and the need for wrapping, is to add categories (sets of extra methods) to the NSArray and NSDictionary classes.

The square brackets don't impair chaining in any way - you can still [[[myDict allKeys] filter:filterMethod] invoke: invokeMethod]

I've tried that, but once you pass in blocks as arguments, the whole thing becomes a lot harder to read. I have no clue how to indent

    [[array filter:^BOOL (id obj) {
            return [obj passesTest];
        }]
        map:^id (id obj) {
            return [obj transformedObject]
        }];

Edit: Also, it's good practice to prefix your categories methods so they don't clash with other category's, so you'd have something like us_map:, us_filter:, reducing readability even further.

It should be a category on NSObject so it supports anything that supports fast enumeration which could be anything. And the indentation in the above is correct. The way you've implemented it is "anti-conventional" and bad form, imho.

Objective-C is a superset of C, functions can be just that - a function. Especially for higher-order functions like map and filter. Not everything needs to be glued to an object.

No shit sherlock.

That's not what he's doing, he's exposing blocks as properties because he couldn't wrap his head around indentation which is the wrong motivation. Maybe read the source before condescending next time.

Uh, I did read it. Yes, it's a bad idea. So is adding random things to NSObject for protocols that it does not conform to. That's actually worse, I'd argue, since you suddenly have -[UIColor map:]. What does that do? Pass RGBA NSNumbers to the block and return a new color? Throw an exception? Silently disregard the message and return nil? At least this approach has type checking.

Just declare a normal function, which can correctly throw a compiler warning if you try to pass anything that does not conform to NSFastEnumeration to it.

The blocks do make it messy but you don't need them, at least in this simple case, if you write your category to take selectors.

[[array filter:@selector(passesTest)] map:@selector(transformedObject)];

That's not perfect, but better.

as long as you don't need additional arguments :)

If indenting gets messy, just put the block in a separate variable. I think it's a shame Xcode doesn't give me an option to put accolades on the same vertical line (Xcode indenting prefers the K&R(?) style).

What's the advantage of implementing map/filter as property accessors of a wrapper object instead of just doing:

    NSArray* map(id<NSFastEnumeration> enumerable, id(^)(id object) block);

Obviously, you'd still want to have a prefix of some sort for namespacing.

In the case of NSDictionary, Underscore.m's map iterates over objects and keys. Fast Enumeration only iterates over the dictionary's keys, when you probably want the value or both.

But I agree, supporting any class that implements NSFastEnumeration would be nice to have.

You may already know this, but you can also enumerate over NSDictionary entries using enumerateKeysAndObjectsUsingBlock: [1] and enumerateKeysAndObjectsWithOptions:usingBlock: [2].

The latter giving you the ability to enumerate in reverse order or spread the enumeration across multiple cores using Grand Central Dispatch.

[1] https://developer.apple.com/library/mac/documentation/Cocoa/...: [2] https://developer.apple.com/library/mac/documentation/Cocoa/...:

It's exactly what's used internally :)

It's cool that we have an Underscore for Obj-C now, but I'm really excited about this because of the project's opinion that bracket syntax sucks.

It's always been a curiosity that a company like Apple, who prides themselves on making great user interfaces can't even get it together to make a programming language that's nice to look at. They seem to care about everyone BUT their own kind these days...

I just spent the last month learning about iOS (and Obj-C). I didn't find the square brackets that terrible. Explicit is better than implicit after all. It seems like dot notation is best used for properties, but I will reserve judgement until I have more than just a month of casual experience.

Indeed. Plus I think well-designed Objective-C is some of the prettiest code out there. Though I've been using it for a few years now almost exclusively (for the most part) so maybe I am a bit biased.

You realize objective-c is almost 30 years old and started as preprocessor for C?

I am not saying that named arguments are a bad idea, in fact they are awesome! Often, they make code a lot more readable, consider for example

    [UIMelon melonWithNumberOfSeeds:10
                  andPriceInDollars:10];

which is a lot more self explanatory than

    new Melon(10, 10)

That being said, for a lot of common and well known fuctions (including map, filter, reduce), Objective-C tends to be overly verbose. I think everybody is looking forward to the new literals, that will allow you to replace

    [NSDictionary dictionaryWithObjectsAndKeys:foo, @"foo",
                                               bar, @"bar",
                                               nil];

with the much more concise

    @{@"foo": foo, @"bar": bar}

Underscore.m tries to achieve something similar. I hope that people will be able see past the - granted, inconventional - syntax to focus on the actual operations done on the data structures. For example:

    // - (NSDictionary *)transformedDictionary;

    NSMutableDictionary *result = [NSMutableDictionary dictionary];
    [myDict enumerateKeysAndObjectsUsingBlock:^(id key, id obj) {
        [result setValue:[obj transformedObject]
                  forKey:key];
    }];
    return result;

vs.

    return _dict(myDict).map(^id (id key, id obj) {
            return [obj transformedObject];
        })
        .unwrap;

>It's cool that we have an Underscore for Obj-C now, but I'm really excited about this because of the project's opinion that bracket syntax sucks. It's always been a curiosity that a company like Apple, who prides themselves on making great user interfaces can't even get it together to make a programming language that's nice to look at. They seem to care about everyone BUT their own kind these days...

Yes, and other people have the uninformed idiotic opition that significant whitespace sucks, or that lots of parenthesis a la Lisp suck.

Alternative syntaxes do not "suck", nor are superficial features like whitespace, brackets and parenthesis the real essence of a syntax (actually those are not even syntactical, they are lexical, and are handled by the lexer, not the parser).

Objective-C is an established language with great features for what it has to do (be dynamic and bridge low level and high level while keeping performance on par with C/C++ when needed). And it has a HUGE API library for both Mobile and Desktop.

Not to mention that it has a noble history of being the language on which both the first web browser and the first modern first person shooter were written in.

Reply on news.ycombinator.com