Hi Greg!

Zelphir,

actually...

Everybody,

i have an awkwardness about not being clear of the meaning of "Library
of Babel".  my sense is it a collection of "subroutines".

in my typical use of Org mode for programming, all my code blocks are
defined in the same .org file, and executed in that file, or tangled to
external files (normally, scripts).  i do use <<noweb>> to include bits
from one part of the same file to the code in another part.  i don't
know if this means i am using the Library of Babel or not.

I love noweb and literate programming facilities. Have done the same thing multiple times, especially, when there was some code I want to learn about and understand or simply explain to my future self. My best example is at: https://notabug.org/ZelphirKaltstahl/the-little-schemer/src/master/chapter-09-y-combinator/code.org Big fan of noweb : )

then, on to...

Zelphir,

i also don't know what is going on.  but, installing geiser,
geiser-guile, and running your code from the source block, looking at
the output results, i get an error that might mean something to you?
see below.  (notice i embedded your call to =org-lob-timediff= at the
end of your source code.)

in addition to with elisp, i use org src blocks with R, shell scripts,
awk, etc.  but, never (till now!) with scheme.

hth.  good luck.  i'll be curious what your next steps are.

cheers, Greg
----

#+HEADER: :var dt1="uninitialized1" dt2="uninitialized2"
#+name: org-lob-timediff
#+begin_src scheme :results output
  (import
   (ice-9 format)
   (srfi srfi-19))


  (define org-timestamp->time-utc
    (λ (timestamp-string)
      (let ([parsed-date (string->date timestamp-string "[~Y-~m-~d ~a ~H:~M]")])
            (date->time-utc parsed-date))))


  (define duration->hours
    (λ (duration)
      ;; 1h = 60min = 3600s
      (/ (time-second duration) 3600)))


  (define org-lob-timediff
    (λ (org-dt1 org-dt2)
      ;; formatting float:

      ;; ~@width, decimals, scale, overflowchar, padchar

      ;; ~ placeholder is following
      ;; @ with sign if negative
      ;; width: minimum width
      ;; decimals: minimum number of digits after decimal point
      ;; scale: ???
      ;; overflowchar: ???
      ;; padchar: char to use for padding

      (format #f
              "~,2f"
              (number->string
               (exact->inexact
                (duration->hours
                 (time-difference (org-timestamp->time-utc org-dt2)
                                  (org-timestamp->time-utc org-dt1))))))))
  (org-lob-timediff "[2021-01-01 Fri 00:00]" "[2021-01-01 Fri 01:45]")
#+end_src

#+RESULTS: org-lob-timediff
: ice-9/boot-9.scm:222:17: In procedure map1:
: Syntax error:
: unknown file:12:0: definition in expression context, where definitions are not allowed, in form (define org-timestamp->time-utc (λ (timestamp-string) (let ((parsed-date (string->date timestamp-string "[~Y-~m-~d ~a ~H:~M]"))) (date->time-utc parsed-date))))
: 
: Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
: scheme@(guile-user) [1]> 

Hmmm this might be a GNU Guile version difference. I remember something about definition context changing in something like Guile 3.0.3 or so, allowing definitions in more places than before. For me this error does not happen, as I use GNU Guile 3.0.7 currently:

~~~~ START ~~~~
scheme@(guile-user)> (version)
$6 = "3.0.7"
~~~~  END  ~~~~

Perhaps org-babel for Guile (I think ob-scheme? or ob-guile?) wraps everything inside another (define ...) and that causes an error with the definition context.

What I did notice before is, that the set locale seems to matter for the short names of weekdays and can lead to an error, but that was on another laptop and not on this system:

~~~~ START ~~~~
executing Scheme code block (org-lob-timediff)...

(dt1 '"uninitialized1")

(dt2 '"uninitialized2")

Starting Geiser REPL ... [3 times]
Guile REPL up and running!
Debug REPL. Enter ,q to quit, ,h for help.
=> ice-9/boot-9.scm:1685:16: In procedure raise-exception:
In procedure string->date: TIME-ERROR type bad-date-format-string: "[~Y-~m-~d ~a ~H:~M]"

Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
scheme@(guile-user) [1]> 
Error during redisplay: (jit-lock-function 315) signaled (wrong-type-argument markerp nil)
Code block evaluation complete.
~~~~  END  ~~~~

But that is a Guile thing and nothing to do with LOB and easily fixed.

I began collecting small snippets and facts about whatever I learn about Emacs in sone repository: https://notabug.org/ZelphirKaltstahl/emacs-usage-guide. Not much there yet, but I will add a working example of LOB, if I can manage to get it working. Perhaps I will also split that in a separate file.

I already got a small example working (not forgetting to run C-c C-v i or `org-babel-lob-ingest` before C-c C-c on the CALL line):

~~~~ START ~~~~
#+HEADER: :var input-str="unspecified"
#+name: org-lob-upcase
#+begin_src scheme
(define upcase
  (λ (str)
    (string-upcase str)))

(upcase input-str)
#+end_src

#+CALL: org-lob-upcase(input-str="test")

#+RESULTS:
: TEST
~~~~  END  ~~~~

So it does not seem to be completely impossible then. The whole syntax-transformer instead of procedure error message thing makes me think, that perhaps it is not using the correct version of Guile or something, so I did the following experiment:

~~~~ START ~~~~
#+name: org-lob-guile-version
#+begin_src scheme
(version)
#+end_src

#+CALL: org-lob-guile-version()

#+RESULTS:
: 3.0.7
~~~~  END  ~~~~

Seems to work fine.

So I still have not understood, what the actual issue is with the timediff function. Perhaps it is something about the imports.

Thanks for taking the time to try and reproduce the behavior!

Best regards,
Zelphir

-- 
repositories: https://notabug.org/ZelphirKaltstahl