A friend of mine drew my attention the welcome message of 4th European Lisp Symposium:
... implementation and application of
any of the Lisp dialects, including
Common Lisp, Scheme, Emacs Lisp,
AutoLisp, ISLISP, Dylan, Clojure,
ACL2, ECMAScript, ...
and then asked if ECMAScript is really a dialect of Lisp. Can it really be considered so? Why?
Is there a well defined and clear-cut set of criteria to help us detect whether a language is a dialect of Lisp? Or is being a dialect taken in a very loose sense (and in that case can we add Python, Perl, Haskell, etc. to the list of Lisp dialects?)
Brendan Eich wanted to do a Scheme-like language for Netscape, but reality intervened and he ended up having to make do with something that looked vaguely like C and Java for "normal" people, but which worked like a functional language.
Personally I think it's an unnecessary stretch to call ECMAScript "Lisp", but to each his own. The key thing about a real Lisp seems like the characteristic that data structure notation and code notation are the same, and that's not true about ECMAScript (or Ruby or Python or any other dynamic functional language that's not Lisp).
Caveat: I have no Lisp credentials :-)
It's not. It's got a lot of functional roots, but so do plenty of other non-lisp languages nowadays, as you pointed out.
Lisps have one remaining characteristic that make them lisps, which is that lisp code is written in terms of lisp data structures (homoiconicity). This is what enables lisps powerful macro system, and why it looks so bizzare to non-lispers. A function call is just a list, where the first element in the list is the name of the function.
Since lisp code is just lisp data, it's possible to do some extremely powerful stuff with metaprogramming, that just can't be done in other languages. Many lisps, even modern ones like clojure, are largely implemented in themselves as a set of macros.
Even though I wouldn't call JavaScript a Lisp, it is, in my humble opinion, more akin to the Lisp way of doing things than most mainstream languages (even functional ones).
For one, just like Lisp, it's, in essence, a simple, imperative language based on the untyped lambda calculus that is fit to be driven by a REPL.
Second, it's easy to embed literal data (including code in the form of lambda expressions) in JavaScript, since a subset of it is equivalent to JSON. This is a common Lisp pattern.
Third, its model of values and types is very lispy. It's object-oriented in a broad sense of the word in that all values have a concept of identity, but it's not particularly object-oriented in most narrower senses of the word. Just as in Lisp, objects are typed and very dynamic. Code is usually split into units of functions, not classes.
In fact, there are a couple of (more or less) recent developments in the JavaScript world that make the language feel pretty lispy at times. Take jQuery, for example. Embedding CSS selectors as a sublanguage is a pretty Lisp-like approach, in my opinion. Or consider ECMAScript Harmony's metaobject protocol: It really looks like a direct port of Common Lisp's (much more so than either Python's or Ruby's metaobject systems!). The list goes on.
JavaScript does lack macros and a sensible implementation of a REPL with editor integration, which is unfortunate. Certainly, influences from other languages are very much visible as well (and not necessarily in a bad way). Still, there is a significant amount of cultural compatibility between the Lisp and JavaScript camps. Some of it may be coincidental (like the recent rise of JavaScript JIT compilation), some systematic, but it's definitely there.
If you call ECMAScript Lisp, you're basically asserting that any dynamic language is Lisp. Since we already have "dynamic language", you're reducing "Lisp" to a useless synonym for it instead of allowing it to have a more specific meaning.
Lisp should properly refer to a language with certain attributes.
A language is Lisp if:
Its source code is tree-structured data, which has a straightforward printed notation as nested lists. Every possible tree structure has a rendering in the corresponding notation and is susceptible to being given a meaning as a construct; the notation itself doesn't have to be extended to extend the language.
The tree-structured data is a principal data structure in the language itself, which makes programs susceptible to manipulation by programs.
The language has symbol data type. Symbols have a printed representation which is interned: when two or more instances of the same printed notation for a symbol appear in the notation, they all denote the same object.
A symbol object's principal virtue is that it is different from all other symbols. Symbols are paired with various other entities in various ways in the semantics of Lisp programs, and thereby serve as names for those entities.
For instance, dialect of Lisp typically have variables, just like other languages. In Lisp, variables are denoted by symbols (the objects in memory) rather than textual names. When part of a Lisp program defines some variable a, the syntax for that a is a symbol object and not the character string "a", which is just that symbol's name for the purposes of printing. A reference to the variable, the expression written as a elsewhere in the program, is also an on object. Because of the way symbols work, it is the same object; this object sameness then connects the reference to the definition. Object sameness might be implemented as pointer equality at the machine level. We know that two symbol values are the same because they are pointers to the same memory location in the heap (an object of symbol type).
Case in point: the NewLisp dialect which has a non-traditional memory management for most data types, including nested lists, makes an exception for symbols by making them behave in the above way. Without this, it wouldn't be Lisp. Quote: "Objects in newLISP (excluding symbols and contexts) are passed by value copy to other user-defined functions. As a result, each newLISP object only requires one reference." [emphasis mine]. Passing symbols too, as by value copy, would destroy their identity: a function receiving a symbol wouldn't be getting the original one, and therefore not correctly receiving its identity.
Compound expressions in a Lisp language—those which are not simple primaries like numbers or strings—consist of a simple list, whose first element is a symbol indicating the operation. The remaining elements, if any, are argument expressions. The Lisp dialect applies some sort of evaluation strategy to reduce the expression to a value, and evoke any side effects it may have.
I would tentatively argue that lists being made of binary cells that hold pairs of values, terminated by a special empty list object, probably should be considered part of the definition of Lisp: the whole business of being able to make a new list out of an existing one by "consing" a new item to the front, and the easy recursion on the "first" and "rest" of a list, and so on.
And then I would stop right there. Some people believe that Lisp systems have to be interactive: provide an environment with a listener, in which everything is mutable, and can be redefined at any time and so on. Some believe that Lisps have to have first-class functions: that there has to be a lambda operator and so on. Staunch traditionalists might even insists that there have to be car and cdr functions, the dotted pair notation supporting improper lists, and that lists have to be made up of cells, and terminated by specifically the symbol nil denoting the empty list, and also a Boolean false. Insisting on car and cdr allows Scheme to be a Lisp, but nil being the list terminator and false rules
The more we shovel into the definition of "Lisp dialect", though, the more it becomes political; people get upset that their favorite dialect (perhaps which they created themselves) is being excluded on some technicality. Insisting on car and cdr allows Scheme to be a Lisp, but nil being the list terminator and false rules it out. What, Scheme not a Lisp?
So, based on the above, ECMAScript isn't a dialect of Lisp. However, an ECMAScript implementation contains functionality which can be exposed as a Lisp dialect and numerous such dialects have been developed. Someone who needs wants ECMAScript to be considered a Lisp for some emotional reasons should perhaps be content with that: that the semantics to support Lisp is there, and just needs a suitable interface to that semantics, which can be developed in ECMAScript and which can interoperate with ECMAScript code.
No it's not.
In order to be considered a Lisp, one has to be homoiconic, which ECMAscript is not.
Not a 'dialect'. I learned LISP in the 70's and haven't used it since, but when I learned JavaScript recently I found myself thinking it was LISP-like. I think that's due to 2 factors: (1) JSON is a list-like associative structures and (2) it's seems as though JS 'objects' are essentially JSON. So even though you don't write JS programs in JSON as you would write LISP in lists, you kind of almost do.
So my answer is that there are enough similarities that programmers familiar with LISP will be reminded of it when they use JavaScript. Statements like JS = LISP in a Java suit are only expressing that feeling. I believe that's all there is to it.
Yes, it is. Quoting Crockford:
"JavaScript has much in common with Scheme. It is a dynamic language. It has a flexible datatype (arrays) that can easily simulate s-expressions. And most importantly, functions are lambdas.
Because of this deep similarity, all of the functions in [recursive programming primer] 'The Little Schemer' can be written in JavaScript."
http://www.crockford.com/javascript/little.html
On the subject of homoiconicity, I would recommend searching that word along with JavaScript. Saying that it is "not homoiconic" is true but not the end of the story.
I think that ECMAScript is a dialect of LISP in the same sense that English is a dialect of French. There are commonalities, but you'll have trouble with assignments in one armed only with knowledge of the other :)
I find it interesting that only one of the three keynote presentations highlighted for the 4th European Lisp Symposium directly concerns Lisp (the other two being about x86/JVM/Python and Scala).
"dialect" is definitely stretching it too far. Still, as someone who has learned and used Python, Javascript, and Scheme, Javascript clearly has a far Lisp-ier feel to it (and Coffeescript probably even more so) than Python.
As for why the European Lisp Symposium would want to portray Javascript as a Lisp, obviously they want to piggyback on the popularity of the Javascript for which the programmer population is many, many times larger than all the rest of the Lisp dialects in their list combined.
Related
I am currently writing a little programming language and have come across a problem. In javascript template literals, we can embed arbitrary expressions, like:
let a = `hello ${ { a: 10, b: 15 } } world`
To properly lex the above-given snippet, the lexer needs to understand bracket matching(parsing essentially) as it can't just assume the first } to be the end of the embedded expression. How do lexers idiomatically solve this problem? One way is to check for proper bracket matching instead of treating them as simple operators, but I am not sure it is the best way. Looking into the code of some javascript lexers also was not very helpful.
The ECMAScript standard specifies (as a theoretical model) different "goal symbols" which are used in different syntactic contexts. Templated strings are one of the contexts with specific goal symbols.
That means that you need a lexical scanner which can switch states. Whether it does so itself, by duplicating part of the work of the parser, or as a result of a syntax action depends on the precise structure of the parsing architecture. You'll find implementations corresponding to both possibilities.
Putting this kind of logic into the parser is easier with predictive (top-down) parsers, such as a recursive descent parser. You could insert a call to the lexer's interface for changing states immediately after recognising the token which triggers the state change (backtick in your templated literal example). Or you could write the lexer interface so the "get a token" function also takes a lexical state argument; then your parser can maintain the lexical state. In effect, this last option is equivalent to using multiple lexical scanners, one for each state, which is also an attractive option, but it depends on separating the lexical scanner from the mechanism for reading input. (Personally, I favour this separation, but it's rarely discussed or implemented.)
Alternatively, you can use a bottom-up parser. In that case you need to be careful with synchronization between the parser's lookahead mechanism and the lexer scanner, since the interface between a bottom-up parser and a lexical scanner always allows the parser to read at least one token ahead, and it's possible that the state change needed to be done before that token was scanned. There are ways to handle this synchronization issue, but it's common for bottom-up parsers to put simple lexical state transitions into the lexer in order to avoid this issue. This necessarily involves a little duplication of effort between scanner and parser but counting braces is not so complicated.
If you're trying to use ECMAScript parsers as a source of inspiration, you need to be aware that there are many other complications with ECMAScript, particularly automatic semicolon insertion, which also involve coordination between lexer and parser. Solving those may impose other constraints on the overall architecture, and certainly makes the resulting parser code harder to read and understand.
I did a search on this topic, but I do not know what keyword to search for so I post this question.
I wondered why the name reduce was chosen in javascript even though the name has been used, such as fold or accumulate, which is more traditional and meaningful(this is my personal opinion).
I've spoken about this topic with someone close to me (one of the people I know who has been dealing with Javascript for a long time and who has also worked with functional languages like Scheme, Racket, and Clojure). He cautiously speculated that this might be an effect of Python.
If you have any traditional context that I do not know, or someone you know about the background of this name, I would be very grateful if you could answer.
If you look at https://en.wikipedia.org/wiki/Fold_(higher-order_function)#Folds_in_various_languages, you can see that the most common names for this operation are fold, reduce, and inject. reduce is by no means an unusual name. In particular, reduce is used in Perl (which JavaScript copied many features from), and Perl probably got it from Lisp (like many other features). The earliest reference I can find is REDUCE in Common Lisp (which was developed in the 1980s and standardized in 1994).
Right to the point, in https://stackoverflow.com/questions/4696618/is-haskell-a-lisp?answertab=votes#tab-top, there is a comment by Kevin Cantu saying:
Yeah, moving beyond the syntax alone, JavaScript is probably more of
a Lisp than Haskell. (Originally conceived as a Scheme
implementation...)
Also, in Lambda the Ultimate: A Lisp to JavaScript Compiler in 100 Lines, they say:
It's immediately quite clear that JS and Lisp have strong ties at the
semantics level [...]
I am familiar with Lisp and functional programming, but not with JavaScript. So these propositions made wonder how powerful is JavaScript. What I've read so far is that it provides lambda expressions and closures. What more functional programming concepts and Lisp-like features does it provide? Does it provide, for instance, tail call recursion, or macros, or ability to manipulate code as data (like Lisp)?
Some things that JavaScript provides that can be considered "Lisp-like":
First class functions (inc. lambdas and closures)
Dynamic typing
Dynamic object model that has some similarities with CLOS
A readable data format that matches the source code format (JSON)
Run time evaluation with an "eval" function (that can be used in an interactive REPL)
Some things that Javascript doesn't have that are pretty common or central to other Lisps:
A homoiconic representation for both code and data (S-expressions)
Built in literals for linked lists / sequences
Support for immutable data structures in general (especially true for Clojure, where every data structure is persistent and immutable)
An extensive macro system for meta-programming
Optional static typing for performance optimisation (e.g. type hints in Common Lisp or Clojure)
Concurrency support
Tail call optimisation
I'm reading the book Structure and Interpretation
of Computer Programs, and I'd like to code a scheme interpreter gradually.
Do you knows the implementation of the scheme most easy to read (and short)?
I will make a JavaScript in C.
SICP itself has several sections detailing how to build a meta-circular interpreter, but I would suggest that you take a look at the following two books for better resources on Scheme interpreters: Programming Languages: Application and Interpretation and Essentials of Programming Languages. They're both easy to read and gradually guide you through building interpreters.
I would recommend the blog series Scheme from scratch which incrementally builds up a scheme interpreter in C.
Christian Queinnec's book Lisp In Small Pieces is superb. More modern that EoPL. Covers both Lisp and Scheme, and goes into detail about the gory low-level stuff that most books omit.
I would recommend reading Kent Dybvig's dissertation "Three Implementation Models for Scheme". Not the whole dissertation, but the first part (up to chapter 3) where he discusses the Heap-Based Model is very suitable for a naive implementation of Scheme.
Another great resource (if I understood it correctly and you want to implement it in C) is Nils Holm's "Scheme 9 from Empty Space". This link is to Nils's page, and there's a link at the bottom to the old, public domain, edition of the book and to the newer, easier to read, commercially available edition. Read both and loved 'em.
I can give you an overview on how my interpreter works, maybe it can give you an idea of the general thing.
Although the answer is pretty late, I hope this can help someone else, who has come to this thread and wants a general idea.
For each line of scheme entered , a Command object is created. If the command is partial then its nest level is stored(number of remaining right brackets to complete the expression). If the command is complete an Expression Object is created and the evaluators are fired on this object.
There are 4 types of evaluator classes defined , each derived from the base class Evaluator
a) Define_Evaluator :for define statements
b) Funcall_Evaluator :for processing other user defined functions
c) Read_Evaluator :for reading an expression and converting it to a scheme object
d) Print_Evaluator :prints the object depending on the type of the object.
e) Eval_Evaluator :does the actual processing of the expression.
3.-> First each expression is read using the Read Evaluator which will create a scheme object out of the expression. Nested expressions are calculated recursively until the expression is complete.
->Next, the Eval_Evaluator is fired which processes the Scheme Expression Object formed in the first step.
this happens as so
a) if the expression to be evaluated is a symbol. Return its value. Therefore the variable blk will return the object for that block.
b) if the expression to be evaluated is a list. Print the list.
c) if the expression to be evaluated is a function. Look for the definition of the function which will return the evaluation using the Funcall_Evaluator.
->Lastly the print evaluator is fired to print the outcome , this print will depend on what type the output expression is.
Disclaimer:
This is how my interpreter works , doesnt have to be that way.
I've been on a similar mission but several years later, recommendations:
Peter Michaux's scheme from scratch: http://michaux.ca/articles/scheme-from-scratch-introduction, and his github repo: https://github.com/petermichaux/bootstrap-scheme/blob/v0.21/scheme.c. Sadly his royal scheme effort seems to have stalled. There were promises of a VM, which with his clarity of explanations would have been great.
Peter Norvigs lis.py: http://norvig.com/lispy.html, although written in python, is very understandable and exploits all the advantages of using a dynamic, weakly typed language to create another. He has a follow up article that adds more advanced features.
Anthony C. Hay used lis.py as an inspiration to create an implementation in C++:
http://howtowriteaprogram.blogspot.co.uk/2010/11/lisp-interpreter-in-90-lines-of-c.html
A more complete implementation is chibi scheme: http://synthcode.com/scheme/chibi/ which does include a VM, but the code base is still not too large to comprehend.
I'm still searching for good blog posts on creating a lisp/scheme VM, which could be coupled with JIT (important for any competitive JS implementation :).
Apart from Queinnec's book, which probably is the most comprehensive one in scheme
to C conversion, you can read also literature from the old platform library.readscheme.org.
When should object literals be used in javascript, sometimes I get confused I am trying to apply oop concepts and pattern to the language. I am trying to not just use procedural programming concepts because I know the language has amazing capabilities.
Object literals are most commonly used as:
a type of associative array; and
a way of passing many arguments to a function.
The second is particularly important and common in libraries like jQuery.
That's how they're commonly used. As for when you should use them, that's a difficult question to answer because it's a bit like asking when should arrays be used. Object literals are a tool. They're a means to an end not an end in itself.
The subtext of your post suggests you're trying to imprint some non-Javascript concepts onto Javascript. I see this a lot (particularly where people try and make everything OO in PHP as the most egregious example). Javascript has its own strengths and weaknesses. Play to those. Don't try to make it something it isn't.
One common mistake is that people confuse OO with Classical language design. You really don't want to be thinking in terms of classes when it comes to javascript, you want to be thinking in terms of functions, duck typing, and prototypes.
This may seem obvious but Javascript object literals are also commonly used for data exchange with other systems. JSON after all is just a subset of Javascript object literal syntax.
Object literals also play an important role in JSON. Though it is important to note JSON is language independent.