veer66

I enjoy using key-value data in dynamic languages. For example, in Python, I can create key-value data for storing the metadata of a document as shown below. I don't discuss why I don't use struct, class, named tuple in this post.

doc_metadata = {"title": "The Rust Programming Language",  
                             "type": "book", 
                             "number-of-pages": 584, 
                             "authors": ["Steve Klabnik", 
                                                "Carol Nichols", 
                                                "contributions"]}

I can code read/write a value easily, for example:

# Write
doc_metadata["type"] = "text book"

# Read
print(doc_metadata["type"])

In Perl and Ruby, we can use Hash, which is almost the same thing as Dict in Python. In JavaScript, we can use an object.

Common Lisp is different. We can use a hash table, but it is not as convenient as Dict in Python.

(let ((doc-metadata (make-hash-table)))
  (setf (gethash :title doc-metadata) "The Rust Programming Language")
  (setf (gethash :type doc-metadata) :BOOK)
  (setf (gethash :number-of-pages doc-metadata) 584)
  (setf (gethash :authors doc-metadata) '("Steve Klabnik"
                                                                      "Carol Nichols" 
                                                                      "contributions")))

Besides construction, printing a hash table is not so convenient. Maybe one can create a function or macro to make creating/printing a hash table convenient. I still felt that I abused Common Lisp.

My code is usually too buggy when I keep mutating the same variable. So I prefer using an immutable data structure to prevent me from messing things up. Moreover, my key-value data usually do not have more than five keys. So I don't strictly need to use an efficient data structure, namely, hash table or binary search tree. So I use alist (assosiation list). I can construct a list like below:

(setq doc-metadata '((:title . "The Rust Programming Language")
                     (:type . :BOOK)
                     (:number-of-pages . 542) 
                     (:authors . '("Steve Klabnik"
                                   "Carol Nichols" 
                                   "contributions"))))

IMO, it looks concise and convenient. We can retrieve key-value pair with a specific key using the assoc function, which I suppose it does linear search. Linear search can be slow. However, my alist doesn't have a lot of keys.

Instead of replacing a value with another value, I can add a new key-value pair with an existing key, for example:

(setq another-doc-metadata (acons :type :TEXT-BOOK doc-metadata))

By retrieving the value of :type using assoc, we get the new value because assoc function retrieves the first key found in alist, for example:

(cdr (assoc :type another-doc-metadata))
;; OUTPUT => :TEXT-BOOK

However, with function calls instead of number/string literal, alist doesn't look concise anymore, for example:

(list (cons :title (get-title x y z))
       (cons :type (get-type x))
       (cons :number-of-pages (get-number-of-pages a b c)) 
       (cons :authors (get-authors c d)))

plist looks much more concise, for example:

(setq doc-metadata (list :title (get-title x y z)
       :type (get-type x)
       :number-of-pages (get-number-of-pages a b c) 
       :authors (get-authors c d)))

I can retrieve a value corresponding to a key easily by getf function. For example:

(getf doc-metadata :type)

A new value can be replaced the old value by setf, example:

(setf (getf doc-mentadata :type) :TEXT-BOOK)

setf is different from acons since acons doesn't mutate the existing list, setf does. Therefore plist is not exactly what I'm looking for.

Maybe the best way is using an Alexandria function for converting plist ot alist as Michał “phoe” Herda suggested.