src/data/data.lisp

   1 ;;; -*- mode: lisp -*-
   2 ;;; Copyright (c) 2005--2007, by A.J. Rossini <blindglobe@gmail.com>
   3 ;;; See COPYRIGHT file for any additional restrictions (BSD license).
   4 ;;; Since 1991, ANSI was finally finished.  Edited for ANSI Common Lisp.
   5
   6 ;;; File:       data.lisp
   7 ;;; Author:     AJ Rossini <blindglobe@gmail.com>
   8 ;;; Copyright:  (c)2007, AJ Rossini.  BSD, LLGPL, or GPLv2, depending
   9 ;;;             on how it arrives.
  10 ;;; Purpose:    data package for lispstat
  11 ;;; Time-stamp: <2006-05-19 12:33:41 rossini>
  12 ;;; Creation:   <2006-05-17 21:34:07 rossini>
  13
  14 ;;; What is this talk of 'release'? Klingons do not make software
  15 ;;; 'releases'.  Our software 'escapes', leaving a bloody trail of
  16 ;;; designers and quality assurance people in its wake.
  17
  18 ;;; This organization and structure is new to the 21st Century
  19 ;;; version.
  20
  21 ;;; conside that data has 3 genotypic chracteristrics.  The first
  22 ;;; would be form -- scalar, vector, array.  second would be datarep
  23 ;;; ("computer science simplistic data" type.  in particular integer,
  24 ;;; real, string, symbol.  The last would be statistical type
  25 ;;; ("usually handled by computer sicience approaches via metadata").
  26 ;;; augmenting datarep type with use in a statistical context,
  27 ;;; i.e. that would include nominal, ordinal, integer, continous,
  28 ;;; interval (orderable subtypes).  Clearly, the statistical type can
  29 ;;; be inherited, likewise the numerical type as well.  The form can
  30 ;;; be pushed up or simplified as necessary, but this can be
  31 ;;; challenging.
  32
  33 ;;; The first approach we will take for CLS is to handle this as
  34 ;;; lisp-only structures.  At the time of realization
  35 ;;; (?instantiation?) of an "abstract" model, the data should be
  36 ;;; pushed into an appropriate form (either "en masse", or
  37 ;;; "on-demand") into a linear algebra framework.
  38
  39 ;;; There is some excellent material on this by John Chambers in one
  40 ;;; of his earlier books.  Reference is being ignored to encourage
  41 ;;; people to read them all.  With all due respect to John, they've
  42 ;;; lasted quite well, but need to be updated.
  43
  44 (in-package :cl-user)
  45
  46 (defpackage :lisp-stat-data
  47   (:documentation "Data management, integration, I/O, and other data technologies.")
  48   (:nicknames :ls-data)
  49   (:use :common-lisp
  50         :lisp-stat-object-system
  51         :lisp-stat-config
  52         :lisp-stat-types
  53         :lisp-stat-compound-data)
  54   (:shadowing-import-from :lisp-stat-object-system
  55                           slot-value call-method call-next-method)
  56   (:export open-file-dialog read-data-file read-data-columns load-data
  57            load-example *variables* *ask-on-redefine*
  58            def variables savevar undef))
  59
  60 (in-package :lisp-stat-data)
  61
  62 ;;; The purpose of this package is to manage data which will be
  63 ;;; processed by LispStat.  In particular, it will be important to
  64 ;;; register variables, datasets, relational structures, and other
  65 ;;; objects which could be the target for statistical modeling and
  66 ;;; inference.
  67
  68 (defvar *lisp-stat-data-table* (make-hash-table)
  69   "Marks up the data the could be used by.")
  70
  71 (defvar *lisp-stat-data-count* 0
  72   "number of items currently recorded.")
  73
  74 ;;; Data (storage) Types, dt-{.*}
  75 ;;;
  76 ;;; Data types are the representation of data from a computer-science
  77 ;;; perspective, i.e. what it is that they contain.  These types
  78 ;;; include particular forms of compound types (i.e. dataframe is
  79 ;;; array-like, but types differ, difference is row-wise, while array
  80 ;;; is a compound of elements of the same type.
  81 ;;;
  82 ;;; This is completely subject to change.
  83
  84 ;;Examples:
  85 ;;  (defun equidimensional (a)
  86 ;;    (or (< (array-rank a) 2)
  87 ;;        (apply #'= (array-dimensions a)))) =>  EQUIDIMENSIONAL
  88 ;;  (deftype square-matrix (&optional type size)
  89 ;;    `(and (array ,type (,size ,size))
  90 ;;          (satisfies equidimensional))) =>  SQUARE-MATRIX
  91
  92 (defun array-of-equal-dt-scalar-type (x)
  93   ;; return dt-scalar-type which fits (more precise that works)
  94   (if x
  95       'integer
  96       nil))
  97
  98 (defun array-of-equal-dt-scalar-type-within-column (x)
  99   ;; return dt-scalar-type which fits (more precise that works)
 100   (if x
 101       'integer
 102       nil))
 103
 104
 105
 106 (deftype dt-scalar (&optional type)
 107   `(or integer double complex symbol))
 108
 109 (deftype dt-array (&optional ndim dimlist)
 110   `(satisfies array-of-equal-dt-scalar-type))
 111
 112 (deftype dt-dataframe (&optional )
 113   `(satisfies array-of-equal-dt-scalar-type-within-column))
 114
 115 ;(deftype dt-relationaldata ()
 116 ;  `(satisfies (foreach unit in relationalUnit
 117 ;              (typep unit 'dt-dataframe))))
 118
 119
 120 ;;; Statistical Variable Types, sv-{.*}
 121 ;;;
 122 ;;; Statistical variable types work to represent the statistical
 123 ;;; category represented by the variable, i.e. nominal, ordinal,
 124 ;;; integral, continous, ratio.   This metadata can be used to hint at
 125 ;;; appropriate analysis methods -- or perhaps more critically, to
 126 ;;; define how these methods will fail in the final interrpretation.
 127
 128 (deftype sv-nominal (&optional n)
 129   ())
 130
 131 (deftype sv-ordinal (ordering &optional n)
 132   ())
 133
 134 (deftype sv-categorical ()
 135   `(satisfies (or sv-nominal sv-ordinal)))
 136 ;;(deftype sv-integer )
 137 ;;(deftype sv-real )  ;; precision could be a secondary component of real, rational, complex.
 138 ;;(deftype sv-rational )
 139 ;;(deftype sv-complex )
 140 ;;(deftype sv-continuous (or 'sv-integer 'sv-real 'sv-rational 'sv-complex)) ;; perhaps, call it "mostly contin..."
 141
 142
 143 ;;; Data I/O
 144
 145 ;; We can read 2 types of data -- those which are pure data, and those
 146 ;; which are impure (lisp-enabled, data as program as data thingy's).
 147
 148 (defparameter *lisp-stat-data-formats*
 149   '(csv tsv))
 150
 151 ;; (defgeneric data-read (srce frmt)
 152 ;;   "read data from stream srce, in format frmt.")
 153
 154 ;; (defgeneric data-write (srce frmt)
 155 ;;   "read data from stream srce, in format frmt.")
 156
 157 ;; (defmacro with-data (body)
 158 ;;   "Stream-handling, maintaining I/O through object typing.")
 159
 160 ;; design-wise should these be replaced with a "with-data" form?
 161
 162 ;;; These need to be elsewhere...!
 163 ;; DSV processing
 164 ;; XML processing
 165
 166 ;;; DM operations should be somewhere else as well.
 167 ;;; Data Management
 168
 169 ;; the goal is to have 2 operations which can be used to create new
 170 ;; data formats out of old ones.
 171
 172 ;; (defgeneric data-subset (ds description)
 173 ;;   "Take a dataset and make it smaller.")
 174
 175 ;; (defgeneric data-relate (ds description)
 176 ;;   "Take 2 or more datasets, and grow them into a bigger one through
 177 ;; relating them (i.e. merge is one example).")
 178
 179 ;;; What should we be able to do?
 180
 181 ;;; Actions on a single dataset
 182 ;;* subset-dataset original-set
 183 ;; :list-of-columns :list-of-rows :list-of-rows-and-columns
 184 ;; :list-of-indices
 185 ;;* resample-dataset original-set ;
 186 ;; :by-rows :by-columns :row-weights :column-weights
 187 ;; :new-number-of-columns :new-number-of-rows
 188 ;;
 189
 190 ;;; Actions based on 2 or more datasets
 191 ;;* concat-dataset set1 set2   ; no matching
 192 ;; :by-row :by-column :kronecker-product
 193 ;;* merge-dataset set1 set2
 194 ;; :match-on-column :match-on-row
 195
 196
 197 ;;; Data tools from "statistics.lsp"
 198
 199 ;;;;
 200 ;;;; Data File Reading
 201 ;;;;
 202
 203 (defun count-file-columns (fname)
 204 "Args: (fname)
 205 Returns the number of lisp items on the first nonblank line of file FNAME."
 206   (with-open-file (f fname)
 207     (if f
 208         (let ((line (do ((line (read-line f) (read-line f)))
 209                         ((or (null line) (< 0 (length line))) line))))
 210           (if line
 211               (with-input-from-string (s line)
 212                 (do ((n 0 (+ n 1)) (eof (gensym)))
 213                     ((eq eof (read s nil eof)) n))))))))
 214
 215 #+xlisp (defvar *xlisptable* *readtable*)
 216
 217 (if (not (fboundp 'open-file-dialog))
 218   #+dialogs
 219   (defun open-file-dialog () ;; why?(&optional set)
 220     (get-string-dialog "Enter a data file name:"))
 221   #-dialogs
 222   (defun open-file-dialog () ;; why? (&optional set)
 223     (error "You must provide a file name explicitly")))
 224
 225 (defun read-data-file (&optional (file (open-file-dialog t)))
 226 "Args:  (file)
 227 Returns a list of all lisp objects in FILE. FILE can be a string or a symbol,
 228 in which case the symbol'f print name is used."
 229   (if file
 230       (let ((eof (gensym)))
 231         (with-open-file (f file)
 232           (if f
 233               (do* ((r (read f nil eof) (read f nil eof))
 234                     (x (list nil))
 235                     (tail x (cdr tail)))
 236                    ((eq r eof) (cdr x))
 237                    (setf (cdr tail) (list r))))))))
 238
 239 ;;; New definition to avoid stack size limit in apply
 240 (defun read-data-columns (&optional (file (open-file-dialog t))
 241                                     (cols (if file
 242                                               (count-file-columns file))))
 243 "Args: (&optional file cols)
 244 Reads the data in FILE as COLS columns and returns a list of lists representing the columns."
 245   (if (and file cols)
 246       (transpose (split-list (read-data-file file) cols))))
 247
 248
 249 ;;; FIXME:AJR:  ALL THE FOLLOWING NEED TO BE SOLVED BY PLATFORM-INDEP PATHNAME WORK!
 250 ;;; FIXME:AJR: use either string or pathname.
 251
 252 (defun path-string-to-path (p s)
 253     (pathname (concatenate 'string (namestring p) s)))
 254
 255 (defun load-data (file)
 256 "Args: (file) as string
 257 Read in data file from the data examples library."
 258   (if (load (path-string-to-path *lispstat-data-dir* file))
 259       t
 260       (load (path-string-to-path *lispstat-examples-dir* file))))
 261
 262 (defun load-example (file)
 263   "Args: (file) as string
 264 Read in lisp example file from the examples library."
 265   (if (load (path-string-to-path *lispstat-examples-dir* file))
 266       t
 267       (load (path-string-to-path *lispstat-data-dir* file))))
 268
 269 ;;;;
 270 ;;;; Listing and Saving Variables and Functions
 271 ;;;;
 272
 273 (defvar *variables* nil)
 274 (defvar *ask-on-redefine* nil)
 275
 276 (defmacro def (symbol value)
 277 "Syntax: (def var form)
 278 VAR is not evaluated and must be a symbol.  Assigns the value of FORM to
 279 VAR and adds VAR to the list *VARIABLES* of def'ed variables. Returns VAR.
 280 If VAR is already bound and the global variable *ASK-ON-REDEFINE*
 281 is not nil then you are asked if you want to redefine the variable."
 282   `(unless (and *ask-on-redefine*
 283                 (boundp ',symbol)
 284                 (not (y-or-n-p "Variable has a value. Redefine?")))
 285     (if (boundp ',symbol)
 286         (setf ,symbol ,value)
 287         (defvar ,symbol ,value))
 288     (pushnew ',symbol *variables*)
 289     ',symbol))
 290
 291 (defun variables-list ()
 292     (mapcar #'intern (sort-data (mapcar #'string *variables*))))
 293
 294 (defun variables ()
 295 "Args:()
 296 Returns a list of the names of all def'ed variables to STREAM"
 297   (if *variables*
 298       (mapcar #'intern (sort-data (mapcar #'string *variables*)))))
 299
 300 (defun savevar (vars file)
 301 "Args: (vars file-name-root)
 302 VARS is a symbol or a list of symbols. FILE-NAME-ROOT is a string (or a symbol
 303 whose print name is used) not endinf in .lsp. The VARS and their current values
 304 are written to the file FILE-NAME-ROOT.lsp in a form suitable for use with the
 305 load command."
 306   (with-open-file (f (concatenate 'string (namestring file) ".lsp")
 307                      :direction :output)
 308     (let ((vars (if (consp vars) vars (list vars))))
 309       (flet ((save-one (x)
 310                (let ((v (symbol-value x)))
 311                  (if (objectp v)
 312                      (format f "(def ~s ~s)~%" x (send v :save))
 313                    (format f "(def ~s '~s)~%" x v)))))
 314             (mapcar #'save-one vars))
 315       vars)))
 316
 317 (defun undef (v)
 318 "Args: (v)
 319 If V is the symbol of a defined variable the variable it is unbound and
 320 removed from the list of defined variables. If V is a list of variable
 321 names each is unbound and removed. Returns V."
 322   (dolist (s (if (listp v) v (list v)))
 323           (when (member s *variables*)
 324                 (setq *variables* (delete s *variables*))
 325                 (makunbound s)))
 326   v)
 327