Re: [racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
On Friday, September 27, 2019 at 2:47:06 PM UTC+2, Matthew Flatt wrote: At Fri, 27 Sep 2019 05:39:23 -0700 (PDT), Jesse Alama wrote: > > On Friday, September 27, 2019 at 2:25:17 PM UTC+2, Matthew Flatt wrote: > > > > > > > > > It looks like `lazy-require` as used by `racket/match` isn't > compatible > > > with creating a new namespace and attaching `racket/match` to that > > > namespace. More generally, `define-runtime-module-path-index` doesn't > > > cooperate with `namespace-attach-module`. Offhand, I don't know how > > > this might be fixed. > > > > > > For your application, I wonder whether you need to create a new > > > namespace at all. The namespace should start out empty, so just > > > `namespace-require` should set it up ok. Do you need to support > > > multiple REPLs or ohterwise have multiple namespaces? > > > > > > > I don't currently need multiple REPLs or namespaces (I think). Using > your > > question and your wondering whether new namespaces are needed as clues, > I > > trimmed the "start a REPL" function down to this: > > > > (define (run-repl) > > (current-read-interaction foo:read-syntax-for-repl) > > (read-eval-print-loop)) > > > > But with this in place I get, in the REPL, trying to evaluate the string > > "$a := 1" (that's the non-S-expression syntax for a simple assignment): > > > > unknown:1:0: #%top-interaction: unbound identifier; > > also, no #%app syntax transformer is bound > > at: #%top-interaction > > in: (#%top-interaction riposte-repl (normal-assignment "a" (expression > > 1))) > > > > Somehow, #%app and #%top-interaction got lost. > > Do you still have `(namespace-require 'foo/expander)` somewhere? That > is still needed to add bindings to the namespace. > > That is, all of the modules in the executable will be declared in the > namespace's registry, but no bindings will have been `require`d into > the namespace. > D'oh. That did the trick. Here's what I've got now: (define (run-repl) (current-read-interaction foo:read-syntax-for-repl) (namespace-require 'foo/expander) (read-eval-print-loop)) That's it. Works like a charm. Thanks! I seemed to have been overthinking things; I thought that I needed a fresh namespace. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/269b4fc1-28f4-4b6e-b93b-d43f7b0c3f69%40googlegroups.com.
Re: [racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
At Fri, 27 Sep 2019 05:39:23 -0700 (PDT), Jesse Alama wrote: > On Friday, September 27, 2019 at 2:25:17 PM UTC+2, Matthew Flatt wrote: > > > > > > It looks like `lazy-require` as used by `racket/match` isn't compatible > > with creating a new namespace and attaching `racket/match` to that > > namespace. More generally, `define-runtime-module-path-index` doesn't > > cooperate with `namespace-attach-module`. Offhand, I don't know how > > this might be fixed. > > > > For your application, I wonder whether you need to create a new > > namespace at all. The namespace should start out empty, so just > > `namespace-require` should set it up ok. Do you need to support > > multiple REPLs or ohterwise have multiple namespaces? > > > > I don't currently need multiple REPLs or namespaces (I think). Using your > question and your wondering whether new namespaces are needed as clues, I > trimmed the "start a REPL" function down to this: > > (define (run-repl) > (current-read-interaction foo:read-syntax-for-repl) > (read-eval-print-loop)) > > But with this in place I get, in the REPL, trying to evaluate the string > "$a := 1" (that's the non-S-expression syntax for a simple assignment): > > unknown:1:0: #%top-interaction: unbound identifier; > also, no #%app syntax transformer is bound > at: #%top-interaction > in: (#%top-interaction riposte-repl (normal-assignment "a" (expression > 1))) > > Somehow, #%app and #%top-interaction got lost. Do you still have `(namespace-require 'foo/expander)` somewhere? That is still needed to add bindings to the namespace. That is, all of the modules in the executable will be declared in the namespace's registry, but no bindings will have been `require`d into the namespace. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/5d8e04c6.1c69fb81.ed990.5373SMTPIN_ADDED_MISSING%40gmr-mx.google.com.
Re: [racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
On Friday, September 27, 2019 at 2:25:17 PM UTC+2, Matthew Flatt wrote: > > > It looks like `lazy-require` as used by `racket/match` isn't compatible > with creating a new namespace and attaching `racket/match` to that > namespace. More generally, `define-runtime-module-path-index` doesn't > cooperate with `namespace-attach-module`. Offhand, I don't know how > this might be fixed. > > For your application, I wonder whether you need to create a new > namespace at all. The namespace should start out empty, so just > `namespace-require` should set it up ok. Do you need to support > multiple REPLs or ohterwise have multiple namespaces? > I don't currently need multiple REPLs or namespaces (I think). Using your question and your wondering whether new namespaces are needed as clues, I trimmed the "start a REPL" function down to this: (define (run-repl) (current-read-interaction foo:read-syntax-for-repl) (read-eval-print-loop)) But with this in place I get, in the REPL, trying to evaluate the string "$a := 1" (that's the non-S-expression syntax for a simple assignment): unknown:1:0: #%top-interaction: unbound identifier; also, no #%app syntax transformer is bound at: #%top-interaction in: (#%top-interaction riposte-repl (normal-assignment "a" (expression 1))) Somehow, #%app and #%top-interaction got lost. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/88108afc-0013-4651-b0c5-aa99e81f5552%40googlegroups.com.
Re: [racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
At Thu, 26 Sep 2019 23:06:28 -0700 (PDT), Jesse Alama wrote: > On Thursday, September 26, 2019 at 3:37:43 PM UTC+2, Matthew Flatt wrote: > Thanks! This helps. I didn't know I was sailing into deep namespace waters. > But another error has reared its head: when, in the REPL, I evaluate an > expression in the foo language, I get > > standard-module-name-resolver: collection not found > for module path: racket/match/gen-match > collection: "racket/match" > in collection directories: >/tmp/build/dist/lib/plt/foo/collects/ It looks like `lazy-require` as used by `racket/match` isn't compatible with creating a new namespace and attaching `racket/match` to that namespace. More generally, `define-runtime-module-path-index` doesn't cooperate with `namespace-attach-module`. Offhand, I don't know how this might be fixed. For your application, I wonder whether you need to create a new namespace at all. The namespace should start out empty, so just `namespace-require` should set it up ok. Do you need to support multiple REPLs or ohterwise have multiple namespaces? -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/5d8dffab.1c69fb81.4b425.5621SMTPIN_ADDED_MISSING%40gmr-mx.google.com.
Re: [racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
On Thursday, September 26, 2019 at 3:37:43 PM UTC+2, Matthew Flatt wrote: At Tue, 24 Sep 2019 22:37:41 -0700 (PDT), Jesse Alama wrote: > > This works for making a standalone executable that can exectute foo > > programs specified on the command line, but doesn't work for a REPL. The > > difficulty seems to be the `namespace-require` part of `run-repl`, > defined > > like this: > > > > > > (define (run-repl) > > (parameterize ([current-namespace (make-base-empty-namespace)]) > > (namespace-require 'foo/expander) > > (read-eval-print-loop))) > > > > > > When I invoke the executable with no arguments, `run-repl` gets called. > But > > this leads to: > > > > > > standard-module-name-resolver: collection not found > > for module path: racket/base/lang/reader > > collection: "racket/base/lang" > > > > The problem is that `racket/base` is declared only in the original > namespace. When you create a fresh namespace, then it gets a fresh > registry, and `racket/base` is not in that registry. > > Really, that goes for `foo/expander`. It's not so much that you want to > refer to `racket/base` as `foo/expander`. While `++lang foo` should > make `foo/expander` be in the original namespace's registry (assuming > that the `foo` language refers to it), it won't be in the fresh > namespace's registry. > > The solution is to attach the module declaration from the original > namespace to the new namespace. > > Here's an example to demonstrate, assuming that the "foo" directory is > installed as a package. The `foo` language here supports just literals > and addition. > > ;; > ;; foo/main.rkt > #lang racket/base > > (provide #%app > #%datum > #%module-begin > #%top-interaction > +) > > (module reader syntax/module-reader > foo) > > ;; > ;; repl.rkt > #lang racket/base > (require racket/cmdline > ;; Ensure that `foo` is here to attach > (only-in foo)) > > (command-line > #:args > ([file #f]) > (cond >[(not file) > (define ns (make-base-empty-namespace)) > (namespace-attach-module (current-namespace) 'foo ns) > (current-namespace ns) > (namespace-require 'foo) > (read-eval-print-loop)] >[else > (dynamic-require file #f)])) > > ;; > ;; Command line > raco exe ++lang foo repl.rkt > raco dist repl-dist repl > Thanks! This helps. I didn't know I was sailing into deep namespace waters. But another error has reared its head: when, in the REPL, I evaluate an expression in the foo language, I get standard-module-name-resolver: collection not found for module path: racket/match/gen-match collection: "racket/match" in collection directories: /tmp/build/dist/lib/plt/foo/collects/ context...: show-collection-err standard-module-name-resolver module-path-index-resolve5 module-path-index-resolve5 namespace-module-use->module+linklet-instances144 for-loop temp37_0 for-loop run-module-instance!125 for-loop loop thunk dynamic-wind loop expand-capturing-lifts loop racket/match is required in (sticking with your filenames) repl.rkt, as well as main.rkt and (not one of your filenames, but essentially just a source of macros that are in turn require'd in main.rkt) expander.rkt. I've added ++lib racket/match to raco exe, to no avail. If I copy all standard Racket collects into the distribution (which makes racket/match in some sense "available"), I get a similar error (possibly the same error, essentially, just with a bit of more information added): no module instance found: # 1 context...: do-error namespace->module-instance70 namespace-module-use->module+linklet-instances144 for-loop temp37_0 for-loop run-module-instance!125 for-loop loop thunk dynamic-wind loop expand-capturing-lifts loop /Users/pltbuild/build/plt-release-64/bundle/racket/collects/racket/repl.rkt:11:26 call-with-values FWIW, the "1" there way at the end of the error message presumably comes from what I'm trying to evaluate in the REPL: (assignment "a" 1) -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/578d1c50-1c95-43ac-8d0e-eb8a97503d7f%40googlegroups.com.
Re: [racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
At Tue, 24 Sep 2019 22:37:41 -0700 (PDT), Jesse Alama wrote: > This works for making a standalone executable that can exectute foo > programs specified on the command line, but doesn't work for a REPL. The > difficulty seems to be the `namespace-require` part of `run-repl`, defined > like this: > > > (define (run-repl) > (parameterize ([current-namespace (make-base-empty-namespace)]) > (namespace-require 'foo/expander) > (read-eval-print-loop))) > > > When I invoke the executable with no arguments, `run-repl` gets called. But > this leads to: > > > standard-module-name-resolver: collection not found > for module path: racket/base/lang/reader > collection: "racket/base/lang" > The problem is that `racket/base` is declared only in the original namespace. When you create a fresh namespace, then it gets a fresh registry, and `racket/base` is not in that registry. Really, that goes for `foo/expander`. It's not so much that you want to refer to `racket/base` as `foo/expander`. While `++lang foo` should make `foo/expander` be in the original namespace's registry (assuming that the `foo` language refers to it), it won't be in the fresh namespace's registry. The solution is to attach the module declaration from the original namespace to the new namespace. Here's an example to demonstrate, assuming that the "foo" directory is installed as a package. The `foo` language here supports just literals and addition. ;; ;; foo/main.rkt #lang racket/base (provide #%app #%datum #%module-begin #%top-interaction +) (module reader syntax/module-reader foo) ;; ;; repl.rkt #lang racket/base (require racket/cmdline ;; Ensure that `foo` is here to attach (only-in foo)) (command-line #:args ([file #f]) (cond [(not file) (define ns (make-base-empty-namespace)) (namespace-attach-module (current-namespace) 'foo ns) (current-namespace ns) (namespace-require 'foo) (read-eval-print-loop)] [else (dynamic-require file #f)])) ;; ;; Command line raco exe ++lang foo repl.rkt raco dist repl-dist repl -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/5d8cbf24.1c69fb81.f4bd9.0e12SMTPIN_ADDED_MISSING%40gmr-mx.google.com.
[racket-users] raco exe & distribute, and namespace-require'ing a module: missing racket/base?
I'm working on building a standalone executable for a #lang that can be
used in two ways:
1. foo awesome.foo: execute file awesome.foo, which is written in #lang foo
2. foo (no arguments): fire up a REPL. Expressions are to be written in the
foo language.
I can get (1) to work, after wrestling with raco exe (using ++lang) and
raco distribute (using ++collects-copy). An invocation that works for me is:
### begin script snippet
raco exe ++lang foo ++lang racket/base ++lang brag foo.rkt
# copy the source code of foo into collects/foo
mkdir -p collects/foo
find . -mindepth 1 -maxdepth 1 -name '*.rkt' -exec cp {} collects/foo ';'
raco distribute ++collects-copy collects dist foo
### end script snippet
This works for making a standalone executable that can exectute foo
programs specified on the command line, but doesn't work for a REPL. The
difficulty seems to be the `namespace-require` part of `run-repl`, defined
like this:
(define (run-repl)
(parameterize ([current-namespace (make-base-empty-namespace)])
(namespace-require 'foo/expander)
(read-eval-print-loop)))
When I invoke the executable with no arguments, `run-repl` gets called. But
this leads to:
standard-module-name-resolver: collection not found
for module path: racket/base/lang/reader
collection: "racket/base/lang"
expander.rkt in foo is written in racket/base, so I suppose that's what
triggers the issue.
My question, crudely put, is: How do I "embed" racket/base so that I can
use `namespace-require` as seen in `run-repl`? (I assume namespace-require
is what I need; if that's wrong, please do correct me.) It appears that
++lang racket/base bit in raco exe isn't enough. A related question about
raco exe is: if racket/base isn't available, what *is* available?
--
You received this message because you are subscribed to the Google Groups
"Racket Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to racket-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/racket-users/4c9c1881-61bb-4e84-a80d-4b3d9e9affd4%40googlegroups.com.
