* bug#36763: `guix search` does unexpected logical and
@ 2019-07-22 17:11 zimoun
2019-07-25 17:35 ` Ludovic Courtès
0 siblings, 1 reply; 6+ messages in thread
From: zimoun @ 2019-07-22 17:11 UTC (permalink / raw)
To: 36763
[-- Attachment #1: Type: text/plain, Size: 808 bytes --]
Dear,
As discussed here [1], the `relevance` in `guix/ui.scm` does not match
"inter-field".
Attached a fix.
Now, the example from the manual
$ guix search crypto library | \
recsel -e '! (name ~ "^(ghc|perl|python|ruby)")' -p name,synopsis
outputs the expected crypto libraries as `libb2`.
Please comment. :-)
Then, please indicate me how the commit has to be filled.
The commit 8874faaaac665100a095ef25e39c9a389f5a397f introducing the
logical AND says:
ui: 'relevance' considers regexps connected with a logical and.
* guix/ui.scm (relevance)[score]: Change to return 0 when one of REGEXPS
doesn't match.
* tests/ui.scm ("package-relevance"): New test.
Should another test be added?
[1] https://lists.gnu.org/archive/html/guix-devel/2019-07/msg00263.html
Thank you in advance,
simon
[-- Attachment #2: search.patch --]
[-- Type: text/x-patch, Size: 2101 bytes --]
diff --git a/guix/ui.scm b/guix/ui.scm
index 7920335928..0e60eb6edc 100644
--- a/guix/ui.scm
+++ b/guix/ui.scm
@@ -1291,23 +1291,36 @@ score, the more relevant OBJ is to REGEXPS."
5 ;exact match
1)))))
regexps))
-
+ scores)
+
+ (define (update relevance weight scores)
+ (map + relevance
+ (map (lambda (score)
+ (* weight score))
+ scores)))
+
+ (let ((scores (fold (lambda (metric relevance)
+ (match metric
+ ((field . weight)
+ (match (field obj)
+ (#f relevance)
+ ((? string? str)
+ (update relevance weight (score str)))
+ ((lst ...)
+ (update relevance weight
+ (fold (lambda (elem prev)
+ (if (zero? (length elem))
+ prev
+ (map + elem prev)))
+ (make-list (length regexps) 0)
+ (map score lst)))
+ )))))
+ (make-list (length regexps) 0)
+ metrics)))
;; Return zero if one of REGEXPS doesn't match.
(if (any zero? scores)
0
- (reduce + 0 scores)))
-
- (fold (lambda (metric relevance)
- (match metric
- ((field . weight)
- (match (field obj)
- (#f relevance)
- ((? string? str)
- (+ relevance (* (score str) weight)))
- ((lst ...)
- (+ relevance (* weight (apply + (map score lst)))))))))
- 0
- metrics))
+ (reduce + 0 scores))))
(define %package-metrics
;; Metrics used to compute the "relevance score" of a package against a set
^ permalink raw reply related [flat|nested] 6+ messages in thread
* bug#36763: `guix search` does unexpected logical and
2019-07-22 17:11 bug#36763: `guix search` does unexpected logical and zimoun
@ 2019-07-25 17:35 ` Ludovic Courtès
2019-09-13 18:23 ` zimoun
0 siblings, 1 reply; 6+ messages in thread
From: Ludovic Courtès @ 2019-07-25 17:35 UTC (permalink / raw)
To: zimoun; +Cc: 36763
Hi Simon,
zimoun <zimon.toutoune@gmail.com> skribis:
> As discussed here [1], the `relevance` in `guix/ui.scm` does not match
> "inter-field".
>
> Attached a fix.
> Now, the example from the manual
> $ guix search crypto library | \
> recsel -e '! (name ~ "^(ghc|perl|python|ruby)")' -p name,synopsis
> outputs the expected crypto libraries as `libb2`.
Nice!
> Then, please indicate me how the commit has to be filled.
> The commit 8874faaaac665100a095ef25e39c9a389f5a397f introducing the
> logical AND says:
>
> ui: 'relevance' considers regexps connected with a logical and.
>
> * guix/ui.scm (relevance)[score]: Change to return 0 when one of REGEXPS
> doesn't match.
> * tests/ui.scm ("package-relevance"): New test.
Yes, you can provide a commit log along these lines.
> Should another test be added?
Yes please, that would be perfect!
> + (define (update relevance weight scores)
> + (map + relevance
> + (map (lambda (score)
> + (* weight score))
> + scores)))
> +
> + (let ((scores (fold (lambda (metric relevance)
> + (match metric
> + ((field . weight)
> + (match (field obj)
> + (#f relevance)
> + ((? string? str)
> + (update relevance weight (score str)))
> + ((lst ...)
> + (update relevance weight
> + (fold (lambda (elem prev)
> + (if (zero? (length elem))
> + prev
> + (map + elem prev)))
> + (make-list (length regexps) 0)
> + (map score lst)))
> + )))))
> + (make-list (length regexps) 0)
> + metrics)))
I don’t have a clear mind on that, but I feel like this could be
simplified somehow. For instance,
(fold p (make-list …) metrics)
looks a lot like:
(map (lambda (x) (fold p x metrics)) regexps)
Well, something like that. :-)
Thanks for working on it!
Ludo’.
^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#36763: `guix search` does unexpected logical and
2019-07-25 17:35 ` Ludovic Courtès
@ 2019-09-13 18:23 ` zimoun
2019-09-16 8:17 ` Ludovic Courtès
0 siblings, 1 reply; 6+ messages in thread
From: zimoun @ 2019-09-13 18:23 UTC (permalink / raw)
To: Ludovic Courtès; +Cc: 36763
[-- Attachment #1: Type: text/plain, Size: 457 bytes --]
Hi Ludo,
Thank you to look at it.
On Thu, 25 Jul 2019 at 19:35, Ludovic Courtès <ludo@gnu.org> wrote:
> (fold p (make-list …) metrics)
>
> looks a lot like:
>
> (map (lambda (x) (fold p x metrics)) regexps)
>
> Well, something like that. :-)
Second try attached with your advice. :-)
I am not sure where to place the newlines on long line.
Then I will try to add tests and send it.
Thank you.
All the best,
simon
[-- Attachment #2: search.patch2 --]
[-- Type: application/octet-stream, Size: 2473 bytes --]
diff --git a/guix/ui.scm b/guix/ui.scm
index 7920335928..80b92bcc5f 100644
--- a/guix/ui.scm
+++ b/guix/ui.scm
@@ -1281,33 +1281,37 @@ weight of this field in the final score.
A score of zero means that OBJ does not match any of REGEXPS. The higher the
score, the more relevant OBJ is to REGEXPS."
- (define (score str)
- (define scores
- (map (lambda (regexp)
- (fold-matches regexp str 0
- (lambda (m score)
- (+ score
- (if (string=? (match:substring m) str)
- 5 ;exact match
- 1)))))
- regexps))
+ (define (score str regexp)
+ (fold-matches regexp str 0
+ (lambda (m score)
+ (+ score
+ (if (string=? (match:substring m) str)
+ 5 ;exact match
+ 1)))))
+
+ (let ((scores (map
+ (lambda (regexp)
+ (fold
+ (lambda (metric relevance)
+ (match metric
+ ((field . weight)
+ (match (field obj)
+ (#f relevance)
+ ((? string? str)
+ (+ relevance (* (score str regexp) weight)))
+ ((lst ...)
+ (+ relevance (* weight
+ (apply + (map
+ (lambda (str)
+ (score str regexp))
+ lst)))))))))
+ 0 metrics))
+ regexps)))
;; Return zero if one of REGEXPS doesn't match.
(if (any zero? scores)
0
- (reduce + 0 scores)))
-
- (fold (lambda (metric relevance)
- (match metric
- ((field . weight)
- (match (field obj)
- (#f relevance)
- ((? string? str)
- (+ relevance (* (score str) weight)))
- ((lst ...)
- (+ relevance (* weight (apply + (map score lst)))))))))
- 0
- metrics))
+ (reduce + 0 scores))))
(define %package-metrics
;; Metrics used to compute the "relevance score" of a package against a set
^ permalink raw reply related [flat|nested] 6+ messages in thread
* bug#36763: `guix search` does unexpected logical and
2019-09-13 18:23 ` zimoun
@ 2019-09-16 8:17 ` Ludovic Courtès
2019-09-19 17:32 ` zimoun
0 siblings, 1 reply; 6+ messages in thread
From: Ludovic Courtès @ 2019-09-16 8:17 UTC (permalink / raw)
To: zimoun; +Cc: 36763
Hello!
zimoun <zimon.toutoune@gmail.com> skribis:
> On Thu, 25 Jul 2019 at 19:35, Ludovic Courtès <ludo@gnu.org> wrote:
>
>> (fold p (make-list …) metrics)
>>
>> looks a lot like:
>>
>> (map (lambda (x) (fold p x metrics)) regexps)
>>
>> Well, something like that. :-)
>
> Second try attached with your advice. :-)
Neat!
> + (let ((scores (map
> + (lambda (regexp)
> + (fold
> + (lambda (metric relevance)
> + (match metric
> + ((field . weight)
> + (match (field obj)
> + (#f relevance)
> + ((? string? str)
> + (+ relevance (* (score str regexp) weight)))
> + ((lst ...)
> + (+ relevance (* weight
> + (apply + (map
> + (lambda (str)
> + (score str regexp))
> + lst)))))))))
> + 0 metrics))
> + regexps)))
For readability, I’d suggest giving a name to one of the two lambdas
above, so you can write, say:
(map regexp-scores regexps)
where:
(define (regexp-scores regexp)
(fold (lambda (metric relevance)
…)
…))
Also, could you add a couple of tests (such as the “libb2” example for
“crypto” + “library” you mentioned earlier)? You can add them to the
existing “package-relevance” test in tests/ui.scm.
Bonus points if you send the patch with ‘git format-patch’ and with a
commit log:
https://guix.gnu.org/manual/en/html_node/Submitting-Patches.html
Thank you!
Ludo’.
^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#36763: `guix search` does unexpected logical and
2019-09-16 8:17 ` Ludovic Courtès
@ 2019-09-19 17:32 ` zimoun
2019-09-19 19:56 ` Ludovic Courtès
0 siblings, 1 reply; 6+ messages in thread
From: zimoun @ 2019-09-19 17:32 UTC (permalink / raw)
To: Ludovic Courtès; +Cc: 36763
Hi Ludo,
On Mon, 16 Sep 2019 at 10:17, Ludovic Courtès <ludo@gnu.org> wrote:
> Bonus points if you send the patch with ‘git format-patch’ and with a
> commit log:
I did my best here:
https://debbugs.gnu.org/cgi/bugreport.cgi?bug=37448
Hope I am doing right.
Thanks,
simon
^ permalink raw reply [flat|nested] 6+ messages in thread
* bug#36763: `guix search` does unexpected logical and
2019-09-19 17:32 ` zimoun
@ 2019-09-19 19:56 ` Ludovic Courtès
0 siblings, 0 replies; 6+ messages in thread
From: Ludovic Courtès @ 2019-09-19 19:56 UTC (permalink / raw)
To: zimoun; +Cc: 36763-done
Hello,
zimoun <zimon.toutoune@gmail.com> skribis:
> On Mon, 16 Sep 2019 at 10:17, Ludovic Courtès <ludo@gnu.org> wrote:
>
>> Bonus points if you send the patch with ‘git format-patch’ and with a
>> commit log:
>
> I did my best here:
> https://debbugs.gnu.org/cgi/bugreport.cgi?bug=37448
>
> Hope I am doing right.
Yup!
We can close it now, thank you.
Ludo’.
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2019-09-19 20:01 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-07-22 17:11 bug#36763: `guix search` does unexpected logical and zimoun
2019-07-25 17:35 ` Ludovic Courtès
2019-09-13 18:23 ` zimoun
2019-09-16 8:17 ` Ludovic Courtès
2019-09-19 17:32 ` zimoun
2019-09-19 19:56 ` Ludovic Courtès
Code repositories for project(s) associated with this public inbox
https://git.savannah.gnu.org/cgit/guix.git
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).