4 ms·
that pictures excite people and draw them in, keep them around They may also excite them because doing pictures the "traditional" way (PHP/GD, GDI, even Cairo)
by gngeal 13y ago
that pictures excite people and draw them in, keep them around
They may also excite them because doing pictures the "traditional" way (PHP/GD, GDI, even Cairo) feels like using stone age tools, while combinator libraries (such as Alex Shinn's fmt, for example) are a hip new thing, for a reason.
Where Racket succeeds over Lisp and Scheme
Huh, that's almost like like saying "Where Visual Studio C++ succeeds over C". It's still on the inside, isn't it? It was a shady move on their part :-), but I guess that whatever helps Racket helps Scheme in general as well.
Nice article, BTW. Thanks. :-)
- moron4hire 13y agoFrom my own reading, it seems that most Schemes are almost 100% R5RS compliant, and don't tend to do much more. When they do much more, it's towards R6RS compliance. R6RS adds a lot, but still nowhere near the full range of a full-package applications development language, much the same as your analogy of VC++ vs, say, C99. And as for Lisp, I guess I'm just used to saying Lisp and meaning Clisp, but any of the "ready for app dev, with parens and prefix notation" languages would fit here--e.g. Clojure--as they all take the approach that compromises in pure syntax are necessary to achieve practical usability. So I think that's where Racket is apart from the Lisps and the Schemes: its core language is syntactically minimalist, as a Scheme, while it strives to have a vast standard library, as a Lisp. Three years ago, they renamed it from PLT Scheme to Racket for this reason. Anyway, I don't know anything about relative user bases for any of the paren languages. I guess "succeeds" was poor word choice on my behalf. I meant more "seems like a better idea to me."
- gnosis 13y ago"So I think that's where Racket is apart from the Lisps and the Schemes: its core language is syntactically minimalist, as a Scheme, while it strives to have a vast standard library, as a Lisp. Three years ago, they renamed it from PLT Scheme to Racket for this reason." Chicken Scheme is in the same position. It has a pretty extensive library, extends far beyond R5RS, avoids R6RS, and plans to be R7RS compliant. It also has the great advantage of compiling down to C.
- irahul 13y ago> Chicken Scheme is in the same position. It has a pretty extensive library, extends far beyond R5RS, May be I am looking from Racket's vantage point, but chicken scheme is R5R7 + some srfi. Compared to Racket which adds oop, myriad of languages(lazy, typed, datalog...), units, factories...chicken looks like a pretty loyal scheme. > It also has the great advantage of compiling down to C. I am curious. If I am not planning cross compilation, what other advantage compiling to C has? I can bundle the runtime to produce a standalone executable in case of racket as well("raco exe" and "raco dist"). Chicken produces smaller executable but that depends on libchicken. Static linking(for distribution, either I will have to bundle libchicken or static link) produces executable big enough that the size difference is negated.
- irahul 13y ago> Huh, that's almost like like saying "Where Visual Studio C++ succeeds over C". It's still on the inside, isn't it? It was a shady move on their part :-) The comparison with Visual C++ isn't apt. Visual C++ for the most part is libraries over C++(not language core). People will complain either ways. When they were PLT Scheme, people cribbed about non-compliance with rnrs. Also, "it's still on the inside" has a threshold. Objective-C and C++ have C inside, but neither of them are C. Racket has added so much to r5rs that it should be considered a scheme derivative, and not a scheme.