load a module without evaluating its body at top level - racket

I have this file:
#lang racket
(provide install global-var)
(define global-var 'test)
(define (install)
(set! global-var '(aaa)))
(install)
And I start the repl, typing so:
Welcome to Racket v5.3.5.
> (define global-var "test007")
> global-var
"test007"
> (require "test.rkt")
'(install)
> global-var
'test
Is it possible to load only some definitions from the file test.rkt, such that the loading does not change the value of global-var, only if I call myself (install), after I load?
I want to use only the standard racket system, not outside packages that are not installed by default in racket.

The short answer is: probably not. You should think of a racket module as a pre-compiled bundle of code that exports certain functions. Any expressions at the top level (such as the call to 'install' here) are conceptually part of the module's setup; allowing a user to use the definitions without running the setup code would invalidate the assumptions of the module writer---not nice for the writer of the module.
To take an example, I have a sound library; requiring the sound library module initializes the sound playback. If there were a tricky way to load the module without running this code, my invariants would be violated.
Perhaps you can explain why it's inconvenient for you to edit the source code?

Related

How to figure out the minimal set of provides from a racket expander module

I am working through the exercises in Matthew Flatt's Creating Languages in Racket. I am currently at step 3, where #lang s-exp "txtadv.rkt top line for a language "client" (world.rkt) is used. In this case, I am assuming that the reader is the s-exp racket reader and that the expander is "txtadv.rkt". Per the text:
"The exports of txtadv.rkt completely determine the bindings that are available in world.rkt—not
only the functions, but also syntactic forms such as require or lambda. For example, txtadv.rkt could
supply a lambda binding to world.rkt that implements a different kind of function than the usual
lambda, such as functions with lazy evaluation."
I see then that txtadv.rkt has this:
(except-out (all-from-out racket) #%module-begin)
I understand by this: re-export (provide) all the imports I got from racket, except for #%module-begin.
What I want: I would like to figure out a proccess to further trim-down this provide list as I do not want to export all of racket.
Things I have tried so far:
I changed at the top of txtadv.rkt to require racket/base and then put this:
(except-out (all-from-out racket/base) #%module-begin)
That worked, but it's still too coarse and fails at the next step with no clue other than revert to require racket.
I started trying to figure out (guess) what s-exps were in "world.rkt" the client, that would require extra syntax:
I commented out the (except-out (all-from-out racket)..)
started "guessing" exports:
that were missing, i.e. define-one-verb
racket functions/forms that are used in macros like symbol->string, list
For all I have tried, the expander txtadv.rkt passes Ctrl-R / syntax-chek, but in the client world.rkt I keep getting:
txtadv.rkt:84:4: define-one-verb: bad syntax in: (define-one-verb north (= n) "go north")

Is it possible to use rackunit in DrRacket with language set to sicp, and if so, how?

I'm starting to work through SICP using DrRacket. I installed the sicp package, and declared #lang sicp at the top of the unit test file and then (require rackunit "xxx.scm"), but I get an unbound identifier error. Is there a mistake, or is it not possible to use rackunit with the sicp package in this way?
You need to use #%require.
#%require is actually a primitive type in the lowest level of racket and it is slightly different than require:
#lang sicp
(#%require rackunit "xxx.scm")
The file you want to test becomes a module so you can use it from other code by providing the identifiers you want to expose:
(#%provide procedure-name)
You can also just require a few needed forms. eg. error and time from racket/base:
(#%require (only racket/base error time))
A hint on finding out where they are is to search the online manuals or from Help > Racket documentation in DrRacket. Eg. here is an example of searching error where you have many choices, but the racket prefixed ones are the ones you're looking for.
NB: Not all forms are compatible across languages. Eg. R5RS has different pair implementation than #lang racket

Portable load/include of define-syntax in R5RS Scheme?

I'm trying to write something that works in both DrRacket/plt-r5rs and Gambit/gsi.
The problem I'm having is that (load "foo.scm") in Gambit does not load define-syntax-blocks. Using (include "foo.scm") in Gambit works, but of course results in a syntax error in DrRacket.
Is there any way to solve this so that I can write portable R5RS code?
Things I've tried:
Redefining (include "foo.scm") to (load "foo.scm") and vice versa. Problem: Illegal to redefine macros in Gambit.
Wrapping said redefinitions in an (if gambit ...). Problem: Illegal to put define inside if(unless inside another define).
Passing string with filename to an include in the library file instead. Problem: Includes in Gambit seem to happen before interpretation starts.
In case it helps: In Racket you can use include in r5rs files:
#lang r5rs
(#%require (only racket include))
(include "foo.scm")
If you define #%require to do nothing in Gambit, then you can use the same source file in both implementations.
It's very hard to write a module that's compatible with both Gambit and Racket.
There are indeed ways you can test for a specific implementation and define things conditionally. There are, in fact, two systems for doing this: SRFI 0 and SRFI 7. Most implementations support one or the other. Not very many support both.
Gambit supports SRFI 0. Racket supports SRFI 7.

What Emacs Lisp function is to `require` as `autoload` is to `load`?

I am trying to program GNU Emacs 23 to issue require commands lazily, on demand, instead of up front in my .emacs file. If I wanted to delay the execution of a load command, I could use autoload. But require and load take different sorts of arguments.
Is there a predefined function that does for require the same job that autoload does for load? And if not, what tools would people recommend that I use to roll my own?
There is not difference between require and load with regards to autoload. require is just a frontend to load, which more or less comes down to:
(defun require (feature &optional filename noerror)
(unless (featurep feature)
(let ((filename (or filename (symbol-name feature))))
(load filename noerror))))
As you can see, the symbol name given to require is equal to the filename given to load. As a matter of fact, the first (require 'foo) evaluated in an Emacs session is equivalent to (load "foo").
Thus, you can just use (auto-load 'foo-function "foo") for foo-function from the library foo, that you can load with (require 'foo).
One more answer to help clarify (this was a little verbose for a comment):
autoload says "if this function isn't already defined, then load this file (if and when the function is called)."
require says "if this library isn't already loaded, then load this file (immediately)."
Note in particular that you don't need to use require to load a library; that's simply the way you ensure that you don't load it again (assuming you don't want to do that). The (provide 'FEATURE) expression in the library will be evaluated no matter how the library was loaded, which lets any future require know that it doesn't need to do anything.
It's a similar situation for autoload -- if the file has already been loaded (and therefore the function in question properly defined), then the autoload no longer has any effect.
What kind of "demand" do you have in mind for your "on demand"?
If a given command or other function needs (or soft-needs) a given library, then that function itself can use (require 'foo) or (require 'foo nil t). The library will be loaded on demand from that function.
Consider also whether you might need to load the file more than once, i.e., reload it in some situations, whether or not it has already been loaded.
For #2, for instance, my code that uses a library of Lisp macros, icicles-mac.el does not just use require, because I want to make sure that if a user gets a new version of that library (e.g., downloads a new source version and byte-compiles it) then that new version is used whenever s?he byte-compiles another library that needs it. (This is important -- when a library of macros changes, other libraries that use those macros generally need to be recompiled after loading the new macros file.) For that, instead of just (require 'icicles-mac) I use this:
(eval-when-compile
(or (condition-case nil
(load-library "icicles-mac") ; Use load-library to ensure latest .elc.
(error nil))
(require 'icicles-mac))) ; Require, so can load separately if not on `load-path'.

How to make and use library with lisp (clisp)?

In C/C++, I can make a library, and make it static one or dll using #include "" in source code, and -labc when linking.
How do I have the same feature in lisp?
As an example of util.lisp in directory A. I define a library function hello.
(defpackage "UTIL"
(:use "COMMON-LISP")
(:nicknames "UT")
(:export "HELLO"))
(in-package util)
(defun hello ()
(format t "hello, world"))
And try to use this library function from main function.
(defun main ()
(ut:hello))
(main)
I tried
clisp main.lisp A/util.lisp
But, I got the following message
*** - READ from #: there is no package with name "UT"
What's the equivalent of #include "" to use the library?
What's the equivalent of -lutil to load the library? What's the command line for clisp/sbcl to use the library?
And for defpackage, Is this equivalent to namespace?
ADDED
I just had to load the library.
(load "./A/util.lisp")
(defun main ()
(ut:hello))
(main)
And run 'clisp main.lisp' works fine.
What you are looking for are called systems. Common Lisp's defpackage has nothing to do with this, and yes, it's about namespaces. Have a look at the HyperSpec, or the idiot's guide (see Xach's comment below) to read more about it.
You can restrict yourself to merely loading files, but usually, a system definition facility is used; mostly ASDF nowadays. A minimal example:
(defsystem my-system
:name "my-system"
:version "0.0.1"
:author "myself"
:license "LLGPL"
:description "it's a system."
:serial t
:components ((:file "packages")
(:file "stuff")
(:file "more_stuff")))
Where packages.lisp would contain the package definition, stuff and more_stuff are the lisp or fasl files to be loaded. This system definition (usually named filename.asd) must be symlinked to (or located in) a directory contained in asdf:*central-registry* for ASDF to find your system. Then, you can load the system thusly:
(asdf:oos 'asdf:load-op 'my-system)
An alternative to this has been added in more recent versions of ASDF:
(asdf:load-system 'my-system)
Or, when using slime, by pressing ,l my-system RET.
You have to load util.lisp before main.lisp:
> (load "util.lisp")
> (load "main.lisp")
> (main)
hello, world
NIL
Practical Common Lisp has a good introduction to defining and using packages.
Common Lisp is an image base language, although usually to a lesser extent than Smalltalk. This means that you use a library by loading it into the image, using LOAD (if used explicitly the often in form (load (compile-file "your-file-here"))), or usually with a system definition facility like ASDF. The loaded code is then available for all code compiled/loaded in the future.
Packages are indeed namespaces. They deal with mapping strings to symbols only, they are not connected directly to files or functions or anything else. You received a package error because you attempted to load a file using a package before a file defining it.