8

GitHub - alphapapa/taxy.el: Programmable taxonomical hierarchy for arbitrary obj...

 3 years ago
source link: https://github.com/alphapapa/taxy.el
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
neoserver,ios ssh client

taxy.el

Now, where did I put that…

This library provides a programmable way to classify arbitrary objects into a hierarchical taxonomy. (That’s a lot of fancy words to say that this lets you put things in nested groups.)

Helpful features include:

Dynamic taxonomiesObjects may be classified into hierarchies automatically defined at runtime based on their attributes. Reusable taxonomiesTaxonomy definitions may be stored in variables and reused in other taxonomies’ descendant groups.

Contents

Examples

May these examples help you classify your understanding.

Numbery (starting basically)

Let’s imagine a silly taxonomy of numbers below 100:

("Numbery" "A silly taxonomy of numbers."
 (("< 10" "Numbers below 10"
   ;; These numbers are leftovers from the sub-taxys below.
   (0 2 4 6 8)
   ;; These sub-taxys further classify the numbers below 10 into odd
   ;; and even.  The odd taxy "consumes" numbers, while the even one
   ;; doesn't, leaving them to reappear in the parent taxy's objects.
   (("Odd" "(consuming)"
     (1 3 5 7 9))
    ("Even" "(non-consuming)"
     (0 2 4 6 8))))
  (">= 10" "Numbers above 9"
   ;; Like in the "< 10" taxy, these numbers are leftovers from this
   ;; taxy's sub-taxys, three of which are non-consuming.
   (10 11 13 14 17 19 22 23 25 26 29 31 34 35 37 38 41 43 46 47 49 50 53 55 58
       59 61 62 65 67 70 71 73 74 77 79 82 83 85 86 89 91 94 95 97 98)
   (("Divisible by 3" "(non-consuming)"
     (12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75 78 81 84
         87 90 93 96 99))
    ("Divisible by 4" "(non-consuming)"
     (12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72 76 80 84 88 92 96))
    ("Divisible by 3 or 4" "(consuming)"
     ;; This taxy consumes numbers it takes in, but since these
     ;; numbers have already been taken in (without being consumed) by
     ;; the previous two sibling taxys, they may also appear in them.
     (12 15 16 18 20 21 24 27 28 30 32 33 36 39 40 42 44 45 48 51 52 54 56 57 60
         63 64 66 68 69 72 75 76 78 80 81 84 87 88 90 92 93 96 99))
    ("Divisible by 5" "(non-consuming)"
     (10 25 35 50 55 65 70 85 95))))))

You might think about how to produce that by writing some imperative code, but taxy allows you to do so in a more declarative and functional manner:

(require 'taxy)

(defvar numbery
  (make-taxy
   :name "Numbery"
   :description "A silly taxonomy of numbers."
   :taxys (list (make-taxy
                 :name "< 10"
                 :description "Numbers below 10 (consuming)"
                 :predicate (lambda (n) (< n 10))
                 :taxys (list
                         ;; These sub-taxys further classify the numbers below 10 into odd
                         ;; and even.  The odd taxy "consumes" numbers, while the even one
                         ;; doesn't, leaving them to reappear in the parent taxy's objects.
                         (make-taxy :name "Odd"
                                    :description "(consuming)"
                                    :predicate #'oddp)
                         (make-taxy :name "Even"
                                    :description "(non-consuming)"
                                    :predicate #'evenp
                                    :then #'identity)))
                (make-taxy
                 :name ">= 10"
                 :description "Numbers above 9 (consuming)"
                 :predicate (lambda (n) (>= n 10))
                 :taxys (list
                         ;; Like in the "< 10" taxy, these sub-taxys further classify
                         ;; the numbers, but only one of them consumes numbers it
                         ;; takes in, leaving the rest to reappear in the parent taxy.
                         (make-taxy :name "Divisible by 3"
                                    :description "(non-consuming)"
                                    :predicate (lambda (n) (zerop (mod n 3)))
                                    :then #'identity)
                         (make-taxy :name "Divisible by 4"
                                    :description "(non-consuming)"
                                    :predicate (lambda (n) (zerop (mod n 4)))
                                    :then #'identity)
                         (make-taxy :name "Divisible by 3 or 4"
                                    :description "(consuming)"
                                    ;; Since this taxy's `:then' function is unset,
                                    ;; it defaults to `ignore', which causes it to
                                    ;; consume numbers it takes in.  Since these
                                    ;; numbers have already been taken in (without
                                    ;; being consumed) by the previous two sibling
                                    ;; taxys, they also appear in them.
                                    :predicate (lambda (n) (or (zerop (mod n 3))
                                                               (zerop (mod n 4)))))
                         (make-taxy :name "Divisible by 5"
                                    :description "(non-consuming)"
                                    :predicate (lambda (n) (zerop (mod n 5)))
                                    :then #'identity))))))

(let ((numbers (cl-loop for i below 100 collect i))
      ;; Since `numbery' is stored in a variable, we use an emptied
      ;; copy of it to avoid mutating the original taxy.
      (taxy (taxy-emptied numbery)))
  (taxy-plain (taxy-fill (reverse numbers) taxy)))

The taxy-fill function applies the numbers in a “cascade” down the hierarchy of “taxys”, and the taxy-plain function returns a meaningful subset of the taxys’ slots, suitable for display.

Lettery (filling incrementally)

You can also add more objects after the hierarchy has been filled:

(defvar lettery
  (make-taxy :name "Lettery"
             :description "A comprehensive taxonomy of letters."
             :taxys (list (make-taxy :name "Vowels"
                                     :description "You know what those are."
                                     :predicate (lambda (l)
                                                  (member-ignore-case l '("a" "e" "i" "o" "u"))))
                          (make-taxy :name "Consonants"
                                     :description "Well, if they aren't a vowel..."))))

(taxy-plain
 (taxy-fill (reverse
             (cl-loop for l from ?a to ?n
                      collect (upcase (char-to-string l))))
            lettery))
("Lettery" "A comprehensive taxonomy of letters."
 (("Vowels" "You know what those are."
   ("A" "E" "I"))
  ("Consonants" "Well, if they aren't a vowel..."
   ("B" "C" "D" "F" "G" "H" "J" "K" "L" "M" "N"))))

Oops, we forgot the letters after N! Let’s add them, too:

(taxy-plain
 (taxy-fill (reverse
             (cl-loop for l from ?n to ?z
                      collect (upcase (char-to-string l))))
            lettery))
("Lettery" "A comprehensive taxonomy of letters."
 (("Vowels" "You know what those are."
   ("O" "U" "A" "E" "I"))
  ("Consonants" "Well, if they aren't a vowel..."
   ("N" "P" "Q" "R" "S" "T" "V" "W" "X" "Y" "Z" "B" "C" "D" "F" "G" "H" "J" "K" "L" "M" "N"))))

Oh, they’re out of order, now. That won’t do. Let’s fix that:

(cl-loop for taxy in-ref (taxy-taxys lettery)
         do (setf (taxy-objects taxy) (cl-sort (taxy-objects taxy) #'<
                                               :key #'string-to-char)))

(taxy-plain lettery)

That’s better:

("Lettery" "A comprehensive taxonomy of letters."
 (("Vowels" "You know what those are."
   ("A" "E" "I" "O" "U"))
  ("Consonants" "Well, if they aren't a vowel..."
   ("B" "C" "D" "F" "G" "H" "J" "K" "L" "M" "N" "N" "P" "Q" "R" "S" "T" "V" "W" "X" "Y" "Z"))))

Sporty (understanding completely)

Let’s try to understand a few things about sports. First we’ll define a struct to make them easier to grasp:

(cl-defstruct sport
  name uses venue fun)

Now we’ll make a list of sports:

(defvar sports
  (list (make-sport :name "Baseball"
                    :uses '(bat ball glove)
                    :venue 'outdoor
                    :fun t)
        (make-sport :name "Football"
                    :uses '(ball)
                    :venue 'outdoor
                    :fun t)
        (make-sport :name "Basketball"
                    :uses '(ball hoop)
                    :venue 'indoor
                    :fun t)
        (make-sport :name "Tennis"
                    :uses '(ball racket)
                    :venue 'outdoor
                    :fun t)
        (make-sport :name "Racquetball"
                    :uses '(ball racket)
                    :venue 'indoor
                    :fun t)
        (make-sport :name "Handball"
                    :uses '(ball glove)
                    :venue 'indoor
                    :fun t)
        (make-sport :name "Soccer"
                    :uses '(ball)
                    :venue 'outdoor
                    :fun nil)
        (make-sport :name "Disc golf"
                    :uses '(disc basket)
                    :venue 'outdoor
                    :fun t)
        (make-sport :name "Ultimate"
                    :uses '(disc)
                    :venue 'outdoor
                    :fun t)
        (make-sport :name "Volleyball"
                    :uses '(ball)
                    :venue 'indoor
                    :fun t)))

And finally we’ll define a taxy to organize them. In this, we use a helper macro to make the member function easier to use in the list of key functions:

(defvar sporty
  (cl-macrolet ((in (needle haystack)
                    `(lambda (object)
                       (when (member ,needle (funcall ,haystack object))
                         ,needle))))
    (make-taxy
     :taxys (list (make-taxy
                   :name "Sporty"
                   :take (lambda (object taxy)
                           (taxy-take-keyed*
                            (list #'sport-venue
                                  (in 'ball 'sport-uses)
                                  (in 'disc 'sport-uses)
                                  (in 'glove 'sport-uses)
                                  (in 'racket 'sport-uses))
                            object taxy
                            ;; We set the `:then' function of the taxys
                            ;; created by `taxy-take-keyed*' to `identity'
                            ;; so they will not consume their objects.
                            :then #'identity)))))))

Now let’s fill the taxy with the sports and format it:

(thread-last sporty
  taxy-emptied
  (taxy-fill sports)
  (taxy-mapcar #'sport-name)
  taxy-plain)
((("Sporty"
   ((disc
     ((outdoor
       ("Ultimate" "Disc golf"))))
    (ball
     ((racket
       ((indoor
         ("Racquetball"))
        (outdoor
         ("Tennis"))))
      (indoor
       ("Volleyball" "Basketball"))
      (outdoor
       ("Soccer" "Football"))
      (glove
       ((indoor
         ("Handball"))
        (outdoor
         ("Baseball"))))))))))

That’s pretty sporty. But classifying them by venue first makes the racket and glove sports not be listed together. Let’s swap that around:

(defvar sporty
  (cl-macrolet ((in (needle haystack)
                    `(lambda (object)
                       (when (member ,needle (funcall ,haystack object))
                         ,needle))))
    (make-taxy
     :taxys (list (make-taxy
                   :name "Sporty"
                   :take (lambda (object taxy)
                           (taxy-take-keyed*
                            (list (in 'ball 'sport-uses)
                                  (in 'disc 'sport-uses)
                                  (in 'glove 'sport-uses)
                                  (in 'racket 'sport-uses)
                                  #'sport-venue)
                            object taxy
                            :then #'identity)))))))

(thread-last sporty
  taxy-emptied
  (taxy-fill sports)
  (taxy-mapcar #'sport-name)
  taxy-plain)
((("Sporty"
   ((disc
     ((outdoor
       ("Ultimate" "Disc golf"))))
    (ball
     ((racket
       ((indoor
         ("Racquetball"))
        (outdoor
         ("Tennis"))))
      (indoor
       ("Volleyball" "Basketball"))
      (outdoor
       ("Soccer" "Football"))
      (glove
       ((indoor
         ("Handball"))
        (outdoor
         ("Baseball"))))))))))

That’s better. But I’d also like to see a very simple classification to help me decide what to play:

(thread-last
    (make-taxy
     :taxys (list
             (make-taxy
              :name "Funny"
              :take (lambda (object taxy)
                      (taxy-take-keyed*
                       (list (lambda (sport)
                               (if (sport-fun sport)
                                   'fun 'boring))
                             #'sport-venue)
                       object taxy)))))
  taxy-emptied
  (taxy-fill sports)
  (taxy-mapcar #'sport-name)
  taxy-plain)
((("Funny"
   ((boring
     ((outdoor
       ("Soccer"))))
    (fun
     ((indoor
       ("Volleyball" "Handball" "Racquetball" "Basketball"))
      (outdoor
       ("Ultimate" "Disc golf" "Tennis" "Football" "Baseball"))))))))

Ah, now I understand.

Applications

Some example applications may be found in the examples directory:

  • Diredy rearranges a Dired buffer into groups by file size and type:
  • Musicy shows a music library with tracks categorized by genre, artist, year, album, etc:

Usage

A taxy is defined with the make-taxy constructor, like:

(make-taxy :name "Numbery"
           :description "A silly taxonomy of numbers."
           :predicate #'numberp
           :then #'ignore
           :taxys (list ...))

The :predicate function determines whether an object fits into that taxy. If it does, taxy-fill adds the object to that taxy’s descendant :taxys, if present, or to its own :objects. The function defaults to identity, so a taxy “takes in” any object by default (i.e. if you only apply objects you want to classify, there’s no need to test them at the top-level taxy).

The :then function determines what happens to an object after being taken in: if the function, called with the object, returns a non-nil value, that value is applied to other taxys at the same level until one of their :then functions returns nil or no more taxys remain. The function defaults to ignore, which makes a taxy “consume” its objects by default. Setting the function to, e.g. identity, makes it not consume them, leaving them eligible to also be taken into subsequent taxys, or to appear in the parent taxy’s objects.

After defining a taxy, call taxy-fill with it and a list of objects to fill the taxy’s hierarchy. Note: taxy-fill modifies the taxy given to it (filling its :objects and those of its :taxys), so when using a statically defined taxy (e.g. one defined with defvar), you should pass taxy-fill a taxy copied with taxy-emptied, which recursively copies a taxy without :objects.

To return a taxy in a more human-readable format (with only relevant fields included), use taxy-plain. You may also use taxy-mapcar to replace objects in a taxy with, e.g. a more useful representation.

Dynamic taxys

You may not always know in advance what taxonomy a set of objects fits into, so taxy lets you add taxys dynamically by using the :take function to add a taxy when an object is “taken into” a parent taxy. For example, you could dynamically classify buffers by their major mode like so:

(defun buffery-major-mode (buffer)
  (buffer-local-value 'major-mode buffer))

(defvar buffery
  (make-taxy
   :name "Buffers"
   :taxys (list
           (make-taxy
            :name "Modes"
            :take (apply-partially #'taxy-take-keyed #'buffery-major-mode)))))

;; Note the use of `taxy-emptied' to avoid mutating the original taxy definition.
(taxy-plain
 (taxy-fill (buffer-list)
            (taxy-emptied buffery)))

The taxy’s :take function is set to the taxy-take-keyed function, partially applied with the buffery-major-mode function as its key-fn (taxy-fill supplies the buffer and the taxy as arguments), and it produces this taxonomy of buffers:

("Buffers"
 (("Modes"
   ((magit-process-mode
     (#<buffer magit-process: taxy.el> #<buffer magit-process: > #<buffer magit-process: notes>))
    (messages-buffer-mode
     (#<buffer *Messages*>))
    (special-mode
     (#<buffer *Warnings*> #<buffer *elfeed-log*>))
    (dired-mode
     (#<buffer ement.el<emacs>>))
    (Custom-mode
     (#<buffer *Customize Apropos*>))
    (fundamental-mode
     (#<buffer  *helm candidates:Bookmarks*> #<buffer *Backtrace*>))
    (magit-diff-mode
     (#<buffer magit-diff: taxy.el> #<buffer magit-diff: notes> #<buffer magit-diff: ement.el>))
    (compilation-mode
     (#<buffer *compilation*> #<buffer *Compile-Log*>))
    (Info-mode
     (#<buffer  *helm info temp buffer*> #<buffer *info*>))
    (help-mode
     (#<buffer *Help*>))
    (emacs-lisp-mode
     (#<buffer ement.el<ement.el>> #<buffer ement-room-list.el> #<buffer *scratch*>
               #<buffer ement-room.el> #<buffer init.el> #<buffer bufler.el>
               #<buffer dash.el> #<buffer *Pp Eval Output*> #<buffer taxy.el> #<buffer scratch.el>))))))

Multi-level dynamic taxys

Of course, the point of taxonomies is that they aren’t restricted to a single level of depth, so you may also use the function taxy-take-keyed* (notice the *) to dynamically make multi-level taxys.

Expanding on the previous example, we use cl-labels to define functions which are used in the taxy’s definition, which are used in the :take function, which calls taxy-take-keyed* (rather than using apply-partially like in the previous example, we use a lambda function, which performs better than partially applied functions). Then when the taxy is filled, a multi-level hierarchy is created dynamically, organizing buffers first by their directory, and then by mode in each directory.

(defvar buffery
  (cl-labels ((buffer-mode (buffer) (buffer-local-value 'major-mode buffer))
              (buffer-directory (buffer) (buffer-local-value 'default-directory buffer)))
    (make-taxy
     :name "Buffers"
     :taxys (list
             (make-taxy
              :name "Directories"
              :take (lambda (object taxy)
                      (taxy-take-keyed* (list #'buffer-directory #'buffer-mode) object taxy)))))))

(taxy-plain
 (taxy-fill (buffer-list)
            (taxy-emptied buffery)))

That produces a list like:

("Buffers"
 (("Directories"
   (("~/src/emacs/ement.el/"
     ((dired-mode
       (#<buffer ement.el<emacs>))
      (emacs-lisp-mode
       (#<buffer ement.el<ement.el> #<buffer ement-room-list.el> #<buffer ement-room.el>))
      (magit-diff-mode
       (#<buffer magit-diff: ement.el>))))
    ("~/src/emacs/taxy.el/"
     ((dired-mode
       (#<buffer taxy.el<emacs>))
      (Info-mode
       (#<buffer *info*>))
      (magit-status-mode
       (#<buffer magit: taxy.el>))
      (emacs-lisp-mode
       (#<buffer taxy-magit-section.el> #<buffer taxy.el<taxy.el> #<buffer scratch.el>))))))))

Reusable taxys

Since taxys are structs, they may be stored in variables and used in other structs (being sure to copy the root taxy with taxy-emptied before filling). For example, this shows using taxy to classify Matrix rooms in Ement.el:

(defun ement-roomy-buffer (room)
  (alist-get 'buffer (ement-room-local room)))

(defvar ement-roomy-unread
  (make-taxy :name "Unread"
             :predicate (lambda (room)
                          (buffer-modified-p (ement-roomy-buffer room)))))

(defvar ement-roomy-opened
  (make-taxy :name "Opened"
             :description "Rooms with buffers"
             :predicate #'ement-roomy-buffer
             :taxys (list ement-roomy-unread
                          (make-taxy))))

(defvar ement-roomy-closed
  (make-taxy :name "Closed"
             :description "Rooms without buffers"
             :predicate (lambda (room)
                          (not (ement-roomy-buffer room)))))

(defvar ement-roomy
  (make-taxy
   :name "Ement Rooms"
   :taxys (list (make-taxy
                 :name "Direct"
                 :description "Direct messaging rooms"
                 :predicate (lambda (room)
                              (ement-room--direct-p room ement-session))
                 :taxys (list ement-roomy-opened
                              ement-roomy-closed))
                (make-taxy
                 :name "Non-direct"
                 :description "Group chat rooms"
                 :taxys (list ement-roomy-opened
                              ement-roomy-closed)))))

Note how the taxys defined in the first three variables are used in subsequent taxys. As well, the ement-roomy-opened taxy has an “anonymous” taxy, which collects any rooms that aren’t collected by its sibling taxy (otherwise those objects would be collected into the parent, “Opened” taxy, which may not always be the most useful way to present the objects).

Using those defined taxys, we then fill the ement-roomy taxy with all of the rooms in the user’s session, and then use taxy-mapcar to replace the room structs with useful representations for display:

(taxy-plain
 (taxy-mapcar (lambda (room)
                (list (ement-room--room-display-name room)
                      (ement-room-id room)))
   (taxy-fill (ement-session-rooms ement-session)
              (taxy-emptied ement-roomy))))

This produces:

("Ement Rooms"
 (("Direct" "Direct messaging rooms"
   (("Opened" "Rooms with buffers"
     (("Unread"
       (("Lars Ingebrigtsen" "!nope:gnus.org")))))
    ("Closed" "Rooms without buffers"
     (("John Wiegley" "!not-really:newartisans.com")
      ("Eli Zaretskii" "!im-afraid-not:gnu.org")))))
  ("Non-direct" "Group chat rooms"
   (("Opened" "Rooms with buffers"
     (("Unread"
       (("Emacs" "!WfZsmtnxbxTdoYPkaT:greyface.org")
        ("#emacs" "!KuaCUVGoCiunYyKEpm:libera.chat")))
      ;; The non-unread buffers in the "anonymous" taxy.
      ((("magit/magit" "!HZYimOcmEAsAxOcgpE:gitter.im")
        ("Ement.el" "!NicAJNwJawmHrEhqZs:matrix.org")
        ("#emacsconf" "!UjTTDnYmSAslLTtMCF:libera.chat")
        ("Emacs Matrix Client" "!ZrZoyXEyFrzcBZKNis:matrix.org")
        ("org-mode" "!rUhEinythPhVTdddsb:matrix.org")
        ("This Week in Matrix (TWIM)" "!xYvNcQPhnkrdUmYczI:matrix.org")))))
    ("Closed" "Rooms without buffers"
     (("#matrix-spec" "!NasysSDfxKxZBzJJoE:matrix.org")
      ("#commonlisp" "!IiGsrmKRHzpupHRaKS:libera.chat")
      ("Matrix HQ" "!OGEhHVWSdvArJzumhm:matrix.org")
      ("#lisp" "!czLxhhEegTEGNKUBgo:libera.chat")
      ("Emacs" "!gLamGIXTWBaDFfhEeO:matrix.org")
      ("#matrix-dev:matrix.org" "!jxlRxnrZCsjpjDubDX:matrix.org")))))))

Threading macros

If you happen to like macros, taxy works well with threading (i.e. thread-last or ->>):

(thread-last ement-roomy
  taxy-emptied
  (taxy-fill (ement-session-rooms ement-session))
  (taxy-mapcar (lambda (room)
                 (list (ement-room--room-display-name room)
                       (ement-room-id room))))
  taxy-plain)

Modifying filled taxys

Sometimes it’s necessary to modify a taxy after filling it with objects, e.g. to sort the objects and/or the sub-taxys. For this, use the function taxy-mapc-taxys (a.k.a. taxy-mapc*). For example, in the sample application musicy.el, the taxys and their objects are sorted after filling, like so:

(defun musicy-files (files)
  (thread-last musicy-taxy
    taxy-emptied
    (taxy-fill files)
    (taxy-mapc* (lambda (taxy)
                  ;; Sort sub-taxys by their name.
                  (setf (taxy-taxys taxy)
                        (cl-sort (taxy-taxys taxy) #'string<
                                 :key #'taxy-name))
                  ;; Sort sub-taxys' objects by name.
                  (setf (taxy-objects taxy)
                        (cl-sort (taxy-objects taxy) #'string<))))
    taxy-magit-section-pp))

Magit section

Showing a taxy with magit-section is very easy:

(require 'taxy-magit-section)

;; Using the `numbery' taxy defined in earlier examples:
(thread-last numbery
  taxy-emptied ;; Get an empty copy of the taxy, since it's defined in a variable.
  (taxy-fill (reverse (cl-loop for i below 30 collect i)))
  taxy-magit-section-pp)

That shows a buffer like this:

Note that taxy-magit-section.el is not installed with the taxy package by default.

Changelog

0.1-pre

Not yet tagged.

Development

Bug reports, feature requests, suggestions — oh my!

License

GPLv3


Recommend

About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK