Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
RSS Feed

Lisp Project of the Day

data-table

You can support this project by donating at:

Donate using PatreonDonate using Liberapay

Or see the list of project sponsors.

data-tabledata-structures

Documentation🥺
Docstrings🤨
Tests 😀
Examples🥺
RepositoryActivity🥺
CI 🥺

This library provides a few methods to work with data tables. Think about it as a limited version of Pandas Dataframe for Common Lisp.

Data-table abilities are really modest. For example, it does not provide methods to investigate data in the table.

To overcome this limitation, we'll use another library - cl-ascii-table and define a describe-object method for data tables:

POFTHEDAY> (defparameter *dt*
             (make-instance 'data-table:data-table
                            :column-names '("Name" "Birthday" "Age")
                            :rows '(("Bob" "1985-05-17" 35)
                                    ("Alice" "1989-02-23" 31)
                                    ("John" "2000-01-03" 20))))

;; This is what we'll wee by default:
POFTHEDAY> (describe *dt*)
#<DATA-TABLE:DATA-TABLE {1003D50293}>
  [standard-object]

Slots with :INSTANCE allocation:
  COLUMN-NAMES                   = ("Name" "Birthday" "Age")
  COLUMN-TYPES                   = (STRING STRING (INTEGER 0 4611686018427387903))
  ROWS                           = (("Bob" "1985-05-17" 35) ("Alice" "1989-02-23" 31))

;; But with custom method we might make it more
;; more convenient:
POFTHEDAY> (defmethod describe-object ((obj data-table:data-table) stream)
             (loop with table = (ascii-table:make-table
                                 (data-table:column-names obj))
                   for row in (data-table:rows obj)
                   do (ascii-table:add-row table row)
                   finally (ascii-table:display table)))

POFTHEDAY> (describe *dt*)
+-------+------------+-----+
| Name  | Birthday   | Age |
+-------+------------+-----+
| Bob   | 1985-05-17 |  35 |
| Alice | 1989-02-23 |  31 |
| John  | 2000-01-03 |  20 |
+-------+------------+-----+

Now let's see what capabilities for data slicing does it have.

Here is how we can retrieve columns. Pay attention, column names are case insensitive. But if case mismatch, column names of the resulting table will be different. Maybe this is a bug:

POFTHEDAY> (describe
            (data-table:select-columns *dt*
                                       '("Name" "Age")))
+-------+-----+
| Name  | Age |
+-------+-----+
| Bob   |  35 |
| Alice |  31 |
| John  |  20 |
+-------+-----+
; No values
POFTHEDAY> (describe
            (data-table:select-columns *dt*
                                       '("name" "age")))
+-------+-----+
| name  | age |
+-------+-----+
| Bob   |  35 |
| Alice |  31 |
| John  |  20 |
+-------+-----+

There is also a way to select a subtable using indices.

This is how we can select contiguous columns with "Name" and "Age" labels:

;; We have to do call this method first.
;; It will fill a list of datatypes.
;; Without filled list of datatypes, make-sub-table will fail :(

POFTHEDAY> (data-table:coerce-data-table-of-strings-to-types *dt*)

POFTHEDAY> (describe
            (data-table:make-sub-table *dt*
                                       :lci 1 :uci 3))
+------------+-----+
| Birthday   | Age |
+------------+-----+
| 1985-05-17 |  35 |
| 1989-02-23 |  31 |
| 2000-01-03 |  20 |
+------------+-----+

If we want to select only the last two rows, then we have to do some math:

;; In pandas.DataFrame this will be just: dt.tail(2)
;;
POFTHEDAY> (let ((rows-count (data-table:number-of-rows *dt*)))
             (describe
              (data-table:make-sub-table *dt*
                                         :lri (- rows-count 2)
                                         :uri rows-count)))
+-------+------------+-----+
| Name  | Birthday   | Age |
+-------+------------+-----+
| Alice | 1989-02-23 |  31 |
| John  | 2000-01-03 |  20 |
+-------+------------+-----+

Also, we can combine column selection by name with make-sub-table. But this will create an intermediate table.

There is also a method overlay-region to combine two tables. Let's use it to add email for Genry and Mary:

POFTHEDAY> (defparameter *dt*
             (data-table:plists-to-data-table
              '((:id 1 :name "Bob")
                (:id 2 :name "Genry")
                (:id 3 :name "Mary"))))

POFTHEDAY> (describe *dt*)
+----+-------+
| ID | NAME  |
+----+-------+
|  1 | Bob   |
|  2 | Genry |
|  3 | Mary  |
+----+-------+

POFTHEDAY> (data-table:overlay-region
            (data-table:plists-to-data-table
             '((:email "genry@gmail.com")
               (:email "mary@hormail.com")))
            *dt*)
NIL

POFTHEDAY> (describe *dt*)
+------------------+-------+
| ID               | NAME  |
+------------------+-------+
| genry@gmail.com  | Bob   |
| mary@hormail.com | Genry |
|                3 | Mary  |
+------------------+-------+

POFTHEDAY> ;; Oh, this method modified the original table :(((

To fix the issue, we need to recreate an original table and to specify indexes for applied overlay:

POFTHEDAY> (defparameter *dt*
             (data-table:plists-to-data-table
              '((:id 1 :name "Bob")
                (:id 2 :name "Genry")
                (:id 3 :name "Mary"))))

POFTHEDAY> (data-table:overlay-region
            (data-table:plists-to-data-table
             '((:email "genry@gmail.com")
               (:email "mary@hormail.com")))
            *dt*
            :col-idx 2
            :row-idx 1)

;; But new column has no name now :(
POFTHEDAY> (describe *dt*)
+----+-------+------------------+
| ID | NAME  | NIL              |
+----+-------+------------------+
|  1 | Bob   | NIL              |
|  2 | Genry | genry@gmail.com  |
|  3 | Mary  | mary@hormail.com |
+----+-------+------------------+

It would be nice to have support for indices and joins like in Pandas. That is what I expect from the data manipulation library.

We've viewed almost all core functions of the data-table library. There are also data coercion and comparison functions. But I think this library has lots of room for improvement.

Maybe @guicho271828 has plans to build a dataframe library as supplementary to his NumCL?


Brought to you by 40Ants under Creative Commons License