Re: [O] [PATCH] Support time units in est+

[Nicolas Goaziou]

Hello,

Malcolm Matalka <address@hidden> writes:

> Hey, this is my first elisp programming so I'm quite certain this is a
> big hack.  I just stole elements from elsewhere in the file.  I'm hoping
> this is good enough to get accepted then perhaps someone with more taste
> would be able to refactor it to be a bit better.

That's not bad. Thank you.

> Let me know if I need to change anything.
>
> From 1167bd20e042ad2ae3f2712f596d76ad8b230336 Mon Sep 17 00:00:00 2001
> From: orbitz <address@hidden>
> Date: Wed, 18 Jan 2017 21:18:23 +0100
> Subject: [PATCH] org-colview.el: Add support for time units to est+
>
> * lisp/org-colview.el: Add support for time units in est+.  Ranges are 
> interpreted just like non-range estimates.

You need to provide the name of the function being modified:

  * lisp/org-colview.el (org-columns--summary-estimate): Add support...
>
> TINYCHANGE
> ---
> lisp/org-colview.el | 44 ++++++++++++++++++++++++++++++++------------
> 1 file changed, 32 insertions(+), 12 deletions(-)
>
> diff --git a/lisp/org-colview.el b/lisp/org-colview.el
> index 45c71a028..2a5c067ac 100644
> --- a/lisp/org-colview.el
> +++ b/lisp/org-colview.el
> @@ -1288,23 +1288,43 @@ When PRINTF is non-nil, use it to format the result."
>     (/ (apply #'+ (mapcar #'org-columns--age-to-seconds ages))
>        (float (length ages)))))
>  
> -(defun org-columns--summary-estimate (estimates printf)
> +(defun org-columns--summary-estimate (estimates _)
>    "Combine a list of estimates, using mean and variance.
> The mean and variance of the result will be the sum of the means
>  and variances (respectively) of the individual estimates."

Could you expound the docstring? In particular, it should make the
output format predictable. See `org-columns--summary-apply-times' for
example.

>    (let ((mean 0)
> -        (var 0))
> +        (var 0)
> +        (hms-flag nil)
> +        (duration-flag nil))
>      (dolist (e estimates)
> -      (pcase (mapcar #'string-to-number (split-string e "-"))
> -(`(,low ,high)
> - (let ((m (/ (+ low high) 2.0)))
> -   (cl-incf mean m)
> -   (cl-incf var (- (/ (+ (* low low) (* high high)) 2.0) (* m m)))))
> -(`(,value) (cl-incf mean value))))
> -    (let ((sd (sqrt var)))
> -      (format "%s-%s"
> -      (format (or printf "%.0f") (- mean sd))
> -      (format (or printf "%.0f") (+ mean sd))))))
> +      (pcase (split-string e "-")
> +        (`(,low ,high)
> +         (dolist (time (list high low))
> +           (cond
> +            (duration-flag)
> +            ((string-match-p org-columns--duration-re time)
> +             (setq duration-flag t))
> +            (hms-flag)
> +            ((string-match-p "\\`[0-9]+:[0-9]+:[0-9]+\\'" time)
> +             (setq hms-flag t))))
> +         (let* ((low-sec (org-columns--time-to-seconds low))
> +                (high-sec (org-columns--time-to-seconds high))
> +                (m (/ (+ low-sec high-sec) 2.0)))
> +           (cl-incf mean m)
> +           (cl-incf var (- (/ (+ (* low-sec low-sec) (* high-sec high-sec)) 
> 2.0) (* m m)))))
> +        (`(,value) (cl-incf mean (org-columns--time-to-seconds
> value)))))

There is much code duplication with `org-columns--summary-apply-times'.
It would be better to avoid it.

We may want to create a new function turning a list of durations into
numbers (with `org-columns--time-to-seconds) and adding the type of the
expected output, as a symbol, in front of the list. Then both
`org-columns--summary-estimate' and `org-columns--summary-apply-times'
could use it.

It would mean we would process the list of estimates two times, but
I don't think it matters much in practice.

If you have an implementation for that, feel free to include it. Do not
consider it a blocker, tho. We can always factor this out later.

> +    (let* ((sd (sqrt var))
> +           (low-second (truncate (- mean sd)))
> +           (high-second (truncate (+ mean sd)))
> +           (low
> +            (cond (duration-flag (org-minutes-to-clocksum-string (/ 
> low-second 60.0)))
> +                  (hms-flag (format-seconds "%h:%.2m:%.2s" low-second))
> +                  (t (format-seconds "%h:%.2m" low-second))))
> +           (high
> +            (cond (duration-flag (org-minutes-to-clocksum-string (/  
> high-second 60.0)))
> +                  (hms-flag (format-seconds "%h:%.2m:%.2s" high-second))
> +                  (t (format-seconds "%h:%.2m" high-second)))))
> +      (format "%s-%s" low high))))

You don't need to `truncate' the estimates.

I don't think the default output format should be "%h:%.2m". If no
special format is present, I think the traditional "%.0f" can be used.

Also, the duplicate `cond' can be factored out. The whole part above can
be written (including suggestions aforementioned):

    (mapconcat (lambda (time)
                 (cond
                  (duration-flag (org-minutes-to-clocksum-string (/ time 60)))
                  (hms-flag (format-seconds "%h:%.2m:%.2s" time))
                  (t (format "%.0f" time))))
               (let ((sd (sqrt var))) (list (- mean sd) (+ mean sd)))
               "-")

It would be nice to add a bunch of tests for various estimates formats.
`test-org-colview/columns-summary' in "test-org-colview.el" contains
already many examples to boot.

Eventually, could you add an entry about this new feature in ORG-NEWS?


Regards,

-- 
Nicolas Goaziou