SRFI-41 and syntax-case
I've worked with lisps for a couple of years, but am new to the various incarnations of macro systems for Scheme (and am by no means proficient with lisp macros in general). I'm using the "jump right in" method to learn more. I'm interested in bringing more libraries to blackhole-libs, and SRFI-41 (streams) is the first I was interested in working with in my project that I'm developing in Gambit. I've extracted the reference implementation from the SRFI, and have it working (at least in the Gambit interpreter) with just a couple of modifications. But a problem has emerged. SRFI-41 makes use of syntax-rules, which Gambit appears to be OK with. It also has one instance of the more-general syntax-case, which appears to be a problem unless I (load "~~/lib/syntax-case") or use the "-:s" option on the command line (this is documented in the manual -- thank you!). If I distribute the SRFI for use in e.g. blackhole, my understanding is that use of (load ...) is brittle and is Considered Harmful in modules. Is there a good approach to addressing this? Should I attempt to re-write the syntax-case macro with either syntax-rules (I assume syntax-rules is not sufficiently powerful enough for this) or with define-macro (macros are intimidating enough as is; I worry about use of a non-hygienic variety as a newcomer...)? Or, is there a way to require the use of the portable syntax-case implementation when the SRFI-41 module is loaded that doesn't make use of a brittle mechanism like load? I didn't want to progress too much farther without verifying this hasn't already been done or is too difficult a task. I didn't see it in SLIB or Blackhole, and I don't think it was in Snow or the Dumping Grounds, but I might have missed it. It seems that everything else in blackhole has avoided the use of syntax-case, but define-macro has been used in at least one place. Thanks, Rick
Afficher les réponses par date
Syntax-case is required because the macro binds an identifier. Syntax-rules is not powerful enough to specify that operation. Define-macro lacks hygiene and should not be considered. On Tue, Sep 28, 2010 at 1:05 PM, R. P. Dillon <rpdillon@gmail.com> wrote:
I've worked with lisps for a couple of years, but am new to the various incarnations of macro systems for Scheme (and am by no means proficient with lisp macros in general). I'm using the "jump right in" method to learn more.
I'm interested in bringing more libraries to blackhole-libs, and SRFI-41 (streams) is the first I was interested in working with in my project that I'm developing in Gambit. I've extracted the reference implementation from the SRFI, and have it working (at least in the Gambit interpreter) with just a couple of modifications. But a problem has emerged.
SRFI-41 makes use of syntax-rules, which Gambit appears to be OK with. It also has one instance of the more-general syntax-case, which appears to be a problem unless I (load "~~/lib/syntax-case") or use the "-:s" option on the command line (this is documented in the manual -- thank you!). If I distribute the SRFI for use in e.g. blackhole, my understanding is that use of (load ...) is brittle and is Considered Harmful in modules.
Is there a good approach to addressing this? Should I attempt to re-write the syntax-case macro with either syntax-rules (I assume syntax-rules is not sufficiently powerful enough for this) or with define-macro (macros are intimidating enough as is; I worry about use of a non-hygienic variety as a newcomer...)? Or, is there a way to require the use of the portable syntax-case implementation when the SRFI-41 module is loaded that doesn't make use of a brittle mechanism like load?
I didn't want to progress too much farther without verifying this hasn't already been done or is too difficult a task. I didn't see it in SLIB or Blackhole, and I don't think it was in Snow or the Dumping Grounds, but I might have missed it. It seems that everything else in blackhole has avoided the use of syntax-case, but define-macro has been used in at least one place.
Thanks, Rick
_______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
OK, I kind of suspected that was the case (i.e. there was a reason syntax-case was used in the first place). Given that, is there any guidance on how to make a blackhole module dependent on ~~/lib/syntax-case? Is it acceptable to simply use (load "~~/lib/syntax-case")? That's the approach that I'm using now, and I wanted to confirm that it was an acceptable approach. Thanks, Rick On Tue, Sep 28, 2010 at 11:17 AM, Phil Bewig <pbewig@gmail.com> wrote:
Syntax-case is required because the macro binds an identifier. Syntax-rules is not powerful enough to specify that operation. Define-macro lacks hygiene and should not be considered.
On Tue, Sep 28, 2010 at 1:05 PM, R. P. Dillon <rpdillon@gmail.com> wrote:
I've worked with lisps for a couple of years, but am new to the various incarnations of macro systems for Scheme (and am by no means proficient with lisp macros in general). I'm using the "jump right in" method to learn more.
I'm interested in bringing more libraries to blackhole-libs, and SRFI-41 (streams) is the first I was interested in working with in my project that I'm developing in Gambit. I've extracted the reference implementation from the SRFI, and have it working (at least in the Gambit interpreter) with just a couple of modifications. But a problem has emerged.
SRFI-41 makes use of syntax-rules, which Gambit appears to be OK with. It also has one instance of the more-general syntax-case, which appears to be a problem unless I (load "~~/lib/syntax-case") or use the "-:s" option on the command line (this is documented in the manual -- thank you!). If I distribute the SRFI for use in e.g. blackhole, my understanding is that use of (load ...) is brittle and is Considered Harmful in modules.
Is there a good approach to addressing this? Should I attempt to re-write the syntax-case macro with either syntax-rules (I assume syntax-rules is not sufficiently powerful enough for this) or with define-macro (macros are intimidating enough as is; I worry about use of a non-hygienic variety as a newcomer...)? Or, is there a way to require the use of the portable syntax-case implementation when the SRFI-41 module is loaded that doesn't make use of a brittle mechanism like load?
I didn't want to progress too much farther without verifying this hasn't already been done or is too difficult a task. I didn't see it in SLIB or Blackhole, and I don't think it was in Snow or the Dumping Grounds, but I might have missed it. It seems that everything else in blackhole has avoided the use of syntax-case, but define-macro has been used in at least one place.
Thanks, Rick
_______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
I would not recommend using (load "~~/lib/syntax-case") and blackhole at the same time. Both of these systems work by overriding Gambit's macro system and implementing their own. Interaction between these systems will not be pretty. At the moment, Blackhole doesn't have support for syntax-case. It's possible that it might be implemented in the future, but it would require changing a fundamental part of what a macro function is because of the stupid way syntax-case is specified. (It requires a macro to be a function that takes exactly one argument, ignoring the normal way of having macro function "creators" that abstract the macro system implementation, like syntax-rules, sc-macro-transformer, er-macro-transformer). The "blackhole way of doing it" would be to rewrite the macro to use syntactic closures, which is the macro system that blackhole is based on. It is hygienic and is also powerful enough to do everything that syntax-case can do. Using syntactic closures is usually easy and feels somewhat similar to define-macro macros. Googling on the term should bring up info on how to use them. One more thing that I should say is that I have been working on creating a package system for blackhole, which is supposed to replace blackhole-libs. Creating a standard library is a much more ambitious task than blackhole-libs ever was, and I believe that a package system would offer much greater flexibility and also make it easier to share code that isn't appropriate in a standard library. The only libraries that will stay in Blackhole are SRFIs, which will be bundled. So blackhole-libs will become obsolete. So work with SRFIs is not wasted as it's easy to move that into Blackhole itself. Hope this helps. 29 sep 2010 kl. 03.37 skrev R. P. Dillon:
OK, I kind of suspected that was the case (i.e. there was a reason syntax-case was used in the first place). Given that, is there any guidance on how to make a blackhole module dependent on ~~/lib/syntax-case? Is it acceptable to simply use (load "~~/lib/syntax-case")? That's the approach that I'm using now, and I wanted to confirm that it was an acceptable approach.
Thanks, Rick
On Tue, Sep 28, 2010 at 11:17 AM, Phil Bewig <pbewig@gmail.com> wrote: Syntax-case is required because the macro binds an identifier. Syntax-rules is not powerful enough to specify that operation. Define-macro lacks hygiene and should not be considered.
On Tue, Sep 28, 2010 at 1:05 PM, R. P. Dillon <rpdillon@gmail.com> wrote: I've worked with lisps for a couple of years, but am new to the various incarnations of macro systems for Scheme (and am by no means proficient with lisp macros in general). I'm using the "jump right in" method to learn more.
I'm interested in bringing more libraries to blackhole-libs, and SRFI-41 (streams) is the first I was interested in working with in my project that I'm developing in Gambit. I've extracted the reference implementation from the SRFI, and have it working (at least in the Gambit interpreter) with just a couple of modifications. But a problem has emerged.
SRFI-41 makes use of syntax-rules, which Gambit appears to be OK with. It also has one instance of the more-general syntax-case, which appears to be a problem unless I (load "~~/lib/syntax-case") or use the "-:s" option on the command line (this is documented in the manual -- thank you!). If I distribute the SRFI for use in e.g. blackhole, my understanding is that use of (load ...) is brittle and is Considered Harmful in modules.
Is there a good approach to addressing this? Should I attempt to re-write the syntax-case macro with either syntax-rules (I assume syntax-rules is not sufficiently powerful enough for this) or with define-macro (macros are intimidating enough as is; I worry about use of a non-hygienic variety as a newcomer...)? Or, is there a way to require the use of the portable syntax-case implementation when the SRFI-41 module is loaded that doesn't make use of a brittle mechanism like load?
I didn't want to progress too much farther without verifying this hasn't already been done or is too difficult a task. I didn't see it in SLIB or Blackhole, and I don't think it was in Snow or the Dumping Grounds, but I might have missed it. It seems that everything else in blackhole has avoided the use of syntax-case, but define-macro has been used in at least one place.
Thanks, Rick
_______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
_______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
Actually, that help a lot! I've been struggling to figure out the right approach, and found the page on schemewiki about syntactic closures, so this gives me a way learn more while solving the problem. I've surveyed a few implementations of scheme that I wanted to work with, and Gambit seems to be both popular and well-documented. But my one gripe was a lack of the core functional libraries I'm used to using in other languages like Clojure and Scala. Various SRFIs provide these, but it seems quite a few are missing (like SRFI 41) in Gambit. So, as I encounter missing SRFIs, I'd like to try to bring them into blackhole if possible. It's great to hear that you're working on a package system. As I learn more, I'd like to write (or port!) some libraries that could be made available through such a system (http servers and clients, graphics libraries and bindings, etc.) I'm not sure how the effort relates to Snow, but Snow seems quite outdated, so I've been focusing on blackhole instead for the time being. Thanks for the excellent reply. On Wed, Sep 29, 2010 at 1:32 AM, Per Eckerdal <per.eckerdal@gmail.com>wrote:
I would not recommend using (load "~~/lib/syntax-case") and blackhole at the same time. Both of these systems work by overriding Gambit's macro system and implementing their own. Interaction between these systems will not be pretty. At the moment, Blackhole doesn't have support for syntax-case. It's possible that it might be implemented in the future, but it would require changing a fundamental part of what a macro function is because of the stupid way syntax-case is specified. (It requires a macro to be a function that takes exactly one argument, ignoring the normal way of having macro function "creators" that abstract the macro system implementation, like syntax-rules, sc-macro-transformer, er-macro-transformer).
The "blackhole way of doing it" would be to rewrite the macro to use syntactic closures, which is the macro system that blackhole is based on. It is hygienic and is also powerful enough to do everything that syntax-case can do. Using syntactic closures is usually easy and feels somewhat similar to define-macro macros. Googling on the term should bring up info on how to use them.
One more thing that I should say is that I have been working on creating a package system for blackhole, which is supposed to replace blackhole-libs. Creating a standard library is a much more ambitious task than blackhole-libs ever was, and I believe that a package system would offer much greater flexibility and also make it easier to share code that isn't appropriate in a standard library. The only libraries that will stay in Blackhole are SRFIs, which will be bundled. So blackhole-libs will become obsolete. So work with SRFIs is not wasted as it's easy to move that into Blackhole itself.
Hope this helps.
29 sep 2010 kl. 03.37 skrev R. P. Dillon:
OK, I kind of suspected that was the case (i.e. there was a reason syntax-case was used in the first place). Given that, is there any guidance on how to make a blackhole module dependent on ~~/lib/syntax-case? Is it acceptable to simply use (load "~~/lib/syntax-case")? That's the approach that I'm using now, and I wanted to confirm that it was an acceptable approach.
Thanks, Rick
On Tue, Sep 28, 2010 at 11:17 AM, Phil Bewig <pbewig@gmail.com> wrote:
Syntax-case is required because the macro binds an identifier. Syntax-rules is not powerful enough to specify that operation. Define-macro lacks hygiene and should not be considered.
On Tue, Sep 28, 2010 at 1:05 PM, R. P. Dillon <rpdillon@gmail.com> wrote:
I've worked with lisps for a couple of years, but am new to the various incarnations of macro systems for Scheme (and am by no means proficient with lisp macros in general). I'm using the "jump right in" method to learn more.
I'm interested in bringing more libraries to blackhole-libs, and SRFI-41 (streams) is the first I was interested in working with in my project that I'm developing in Gambit. I've extracted the reference implementation from the SRFI, and have it working (at least in the Gambit interpreter) with just a couple of modifications. But a problem has emerged.
SRFI-41 makes use of syntax-rules, which Gambit appears to be OK with. It also has one instance of the more-general syntax-case, which appears to be a problem unless I (load "~~/lib/syntax-case") or use the "-:s" option on the command line (this is documented in the manual -- thank you!). If I distribute the SRFI for use in e.g. blackhole, my understanding is that use of (load ...) is brittle and is Considered Harmful in modules.
Is there a good approach to addressing this? Should I attempt to re-write the syntax-case macro with either syntax-rules (I assume syntax-rules is not sufficiently powerful enough for this) or with define-macro (macros are intimidating enough as is; I worry about use of a non-hygienic variety as a newcomer...)? Or, is there a way to require the use of the portable syntax-case implementation when the SRFI-41 module is loaded that doesn't make use of a brittle mechanism like load?
I didn't want to progress too much farther without verifying this hasn't already been done or is too difficult a task. I didn't see it in SLIB or Blackhole, and I don't think it was in Snow or the Dumping Grounds, but I might have missed it. It seems that everything else in blackhole has avoided the use of syntax-case, but define-macro has been used in at least one place.
Thanks, Rick
_______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
_______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list
participants (3)
-
Per Eckerdal -
Phil Bewig -
R. P. Dillon