4 ms·

How GNU Guile is 10x better (2021)

101 points by Tomte 10mo ago

cdaringe 10mo ago
Definitely some interesting and fun properties. Pretty hard to consider them, in totality, anything close to 10x-ing from the herd. Software is art. Maybe someone out there somehow gets 10xd from these traits, but highly unlikely
- groundzeros2015 10mo ago
  The 10x argument for lisp was more compelling when everyone was using C and Fortran instead of JS and python. The contrasts here do seem pretty small. The real productivity benefit for me is that the ecosystem is so stable and well documented. You can learn it all and keep it in your head.
  - ux266478 10mo ago
    Guile is an embedded scripting and configuration language. Its competition is Lua. Which is unfortunate for Guile, because it's even less attractive there.
    - ykonstant 10mo ago
      What are the pros and cons of Guile wrt Lua?
      - ux266478 10mo ago
        The only pro I can think of is you get a standard scheme set of batteries, which isn't even really a pro in this space. The biggest con is implementation size and complexity. IIRC the JITter alone is twice the size of LuaJIT's entire codebase, and not for being vastly better. License problems too, if you're not making copyleft software. Guile is GPL, both Luas are MIT.
spit2wind 10mo ago
Re: 8 complete info-manual Yes, it has an info manual and, I agree, info is the superior documentation viewer. However, a good browser is no replacement for bad writing. The Guile manual is not well written. The organization seems almost random. The text emphasizes minutia while glossing over fundamental details. It off-loads much to RnRSs and SFRIs (whatever those are). Basically, it suffers badly from The Curse of Expertise. The documentation's shortcomings might be okay except that Guile is, or was, the premier extension language for the whole of the GNU project. I considered trying to improve the manual, but why would I dedicate time and effort to a language that I don't know and whose community can't follow it's own advice? Consider the following: "Make sure your manual is clear to a reader who knows nothing about the topic and reads it straight through. This means covering basic topics at the beginning, and advanced topics only later. This also means defining every specialized term when it is first used." https://www.gnu.org/prep/standards/html_node/GNU-Manuals.html https://www.gnu.org/prep/standards/html_node/GNU-Manuals.htm... Most of these points: https://www.fsf.org/campaigns/gnu-press/GNU-Press-styleguide.pdf https://www.fsf.org/campaigns/gnu-press/GNU-Press-styleguide... Maybe at FOSDEM this year, people could do a Hackathon and knock out some basics, like defining acronyms or using terms only after they're defined. PS: every Python tarball for quite a while has instructions for building the documentation, including in info format
- Qem 10mo ago
  For those with no previous experience with Scheme, how does one learn Guile? Are there recommended books or MOOCs? Is familiarity with Emacs effectively a pre-requisite, as it happens with most open-source lisps?
  - petre 10mo ago
    Try Racket instead, much better documentation.
- throwaway17_17 10mo ago
  In case someone is reading and wonders: RnRS - Revised n Report on Scheme (where 1 <= n <= 7) These are basically Scheme Editions. R5RS, R6RS, and R7RS are the ‘big ones’ that are commonly referenced, R7RS being issued in 2013 (5 — 1998, 6 — 2007). SRFI - Scheme Request for Implementation SFRI is basically an informal standards type document. SFRI’s are typically used to request a common library feature for implementation (more useful before R6RS which essentially introduces a functioning standard library for scheme. Most implementations acknowledge that they implement SFRI #n as a quick reference for what ‘extras’ are in their shipped stdlib. Note that I think parent may have been rhetorically asking, or asking with heavy sarcasm. Also, I agree that the Manual is not written that well. It is pretty big, but if Guile is going to continue playing a role as the ‘Scheme of Record’ in GNU and in Linux more generally, it should meet modern expectations for documentation.
transfire 10mo ago
Honestly if it were not for my extensive Ruby background that I have now been able to carry over to Crystal, I probably would have dived into Guile. (I have been enjoying Elixir too, but at the end of the day it doesn’t quite sit right with me — just feels a bit clunky. Gleam seems an attractive alternative though. The BEAM rocks, but it is a heavy dependency that doesn’t fit all distribution needs.)
zelphirkalt 10mo ago
The existing purely functional data structures in IJP's library guile-pfds are great. However, be aware, that this library has not been maintained for a long time and that attempts to contact IJP have failed. So far I have used them and have not encountered a problem. One funny thing I just noticed: I am not the only one often mistyping "PFDS" as "PDFs" (usually in lowercase though). On IJP's repo for "fectors": > One such implementation is based on fingertrees and is provided as part of my pdfs package[1]

101 points by Tomte 10mo ago

Definitely some interesting and fun properties. Pretty hard to consider them, in totality, anything close to 10x-ing from the herd. Software is art. Maybe someone out there somehow gets 10xd from these traits, but highly unlikely

groundzeros2015 10mo ago
The 10x argument for lisp was more compelling when everyone was using C and Fortran instead of JS and python. The contrasts here do seem pretty small. The real productivity benefit for me is that the ecosystem is so stable and well documented. You can learn it all and keep it in your head.
- ux266478 10mo ago
  Guile is an embedded scripting and configuration language. Its competition is Lua. Which is unfortunate for Guile, because it's even less attractive there.
  - ykonstant 10mo ago
    What are the pros and cons of Guile wrt Lua?
    - ux266478 10mo ago
      The only pro I can think of is you get a standard scheme set of batteries, which isn't even really a pro in this space. The biggest con is implementation size and complexity. IIRC the JITter alone is twice the size of LuaJIT's entire codebase, and not for being vastly better. License problems too, if you're not making copyleft software. Guile is GPL, both Luas are MIT.

spit2wind 10mo ago

Re: 8 complete info-manual Yes, it has an info manual and, I agree, info is the superior documentation viewer. However, a good browser is no replacement for bad writing. The Guile manual is not well written. The organization seems almost random. The text emphasizes minutia while glossing over fundamental details. It off-loads much to RnRSs and SFRIs (whatever those are). Basically, it suffers badly from The Curse of Expertise. The documentation's shortcomings might be okay except that Guile is, or was, the premier extension language for the whole of the GNU project. I considered trying to improve the manual, but why would I dedicate time and effort to a language that I don't know and whose community can't follow it's own advice? Consider the following: "Make sure your manual is clear to a reader who knows nothing about the topic and reads it straight through. This means covering basic topics at the beginning, and advanced topics only later. This also means defining every specialized term when it is first used." https://www.gnu.org/prep/standards/html_node/GNU-Manuals.html https://www.gnu.org/prep/standards/html_node/GNU-Manuals.htm... Most of these points: https://www.fsf.org/campaigns/gnu-press/GNU-Press-styleguide.pdf https://www.fsf.org/campaigns/gnu-press/GNU-Press-styleguide... Maybe at FOSDEM this year, people could do a Hackathon and knock out some basics, like defining acronyms or using terms only after they're defined. PS: every Python tarball for quite a while has instructions for building the documentation, including in info format

Qem 10mo ago
For those with no previous experience with Scheme, how does one learn Guile? Are there recommended books or MOOCs? Is familiarity with Emacs effectively a pre-requisite, as it happens with most open-source lisps?
- petre 10mo ago
  Try Racket instead, much better documentation.
throwaway17_17 10mo ago
In case someone is reading and wonders: RnRS - Revised n Report on Scheme (where 1 <= n <= 7) These are basically Scheme Editions. R5RS, R6RS, and R7RS are the ‘big ones’ that are commonly referenced, R7RS being issued in 2013 (5 — 1998, 6 — 2007). SRFI - Scheme Request for Implementation SFRI is basically an informal standards type document. SFRI’s are typically used to request a common library feature for implementation (more useful before R6RS which essentially introduces a functioning standard library for scheme. Most implementations acknowledge that they implement SFRI #n as a quick reference for what ‘extras’ are in their shipped stdlib. Note that I think parent may have been rhetorically asking, or asking with heavy sarcasm. Also, I agree that the Manual is not written that well. It is pretty big, but if Guile is going to continue playing a role as the ‘Scheme of Record’ in GNU and in Linux more generally, it should meet modern expectations for documentation.

transfire 10mo ago

Honestly if it were not for my extensive Ruby background that I have now been able to carry over to Crystal, I probably would have dived into Guile. (I have been enjoying Elixir too, but at the end of the day it doesn’t quite sit right with me — just feels a bit clunky. Gleam seems an attractive alternative though. The BEAM rocks, but it is a heavy dependency that doesn’t fit all distribution needs.)

zelphirkalt 10mo ago

The existing purely functional data structures in IJP's library guile-pfds are great. However, be aware, that this library has not been maintained for a long time and that attempts to contact IJP have failed. So far I have used them and have not encountered a problem. One funny thing I just noticed: I am not the only one often mistyping "PFDS" as "PDFs" (usually in lowercase though). On IJP's repo for "fectors": > One such implementation is based on fingertrees and is provided as part of my pdfs package[1]