src/data/data-clos.lisp

   1 ;;; -*- mode: lisp -*-
   2
   3 ;;; Time-stamp: <2008-10-03 02:15:49 tony>
   4 ;;; Creation:   <2008-03-12 17:18:42 blindglobe@gmail.com>
   5 ;;; File:       data-clos.lisp
   6 ;;; Author:     AJ Rossini <blindglobe@gmail.com>
   7 ;;; Copyright:  (c)2008, AJ Rossini.  BSD, LLGPL, or GPLv2, depending
   8 ;;;             on how it arrives.
   9 ;;; Purpose:    data package for lispstat.   redoing data structures
  10 ;;;             in a CLOS based framework.
  11 ;;;
  12 ;;; No real basis for work, there is a bit of new-ness and R-ness to
  13 ;;; this work. In particular, the notion of relation is key and
  14 ;;; integral to the analysis.  Tables are related and matched
  15 ;;; vectors,for example.  "column" vectors are related observations
  16 ;;; (by measure/recording) while "row" vectors are related readings
  17 ;;; (by case)
  18
  19
  20 ;;; What is this talk of 'release'? Klingons do not make software
  21 ;;; 'releases'.  Our software 'escapes', leaving a bloody trail of
  22 ;;; designers and quality assurance people in its wake.
  23
  24 ;;; This organization and structure is new to the 21st Century
  25 ;;; version.
  26
  27 (in-package :lisp-stat-data-clos)
  28
  29 ;;; Relational structure -- can we capture a completely unnormalized
  30 ;;; data strucutre to propose possible modeling approaches, and
  31 ;;; propose appropriate models and inferential strategies?
  32
  33 ;; verb-driven schema for data collection.  Should encode independence
  34 ;; or lack of when possible.
  35
  36 #+nil(progn
  37        (def-statschema MyDB
  38            :tables (list (list t1 )
  39                          (list t2 )
  40                          (list t4 ))
  41            :unique-key key
  42            :stat-relation '(t1 (:nest-within t2) (:nest-within t3))
  43            :))
  44
  45 ;; Need to figure out typed vectors.   We then map a series of typed
  46 ;; vectors over to tables where columns are equal typed.  In a sense,
  47 ;; this is a relation (1-1) of equal-typed arrays.  For the most part,
  48 ;; this ends up making the R data.frame into a relational building
  49 ;; block (considering 1-1 mappings using row ID as a relation).
  50 ;; Is this a worthwhile generalization?
  51
  52 ;;; verbs vs semantics for dt conversion -- consider the possibily of
  53 ;;; how adverbs and verbs relate, where to put which semantically to
  54 ;;; allow for general approach.
  55
  56 ;;; eg. Kasper's talk on the FUSION collection of parsers.
  57
  58
  59 ;;; Need to consider modification APIs
  60 ;;; actions are:
  61 ;;; - import
  62 ;;; - get/set row names (case names)
  63 ;;; -         column names (variable names)
  64 ;;; -         dataset values
  65 ;;; -         annotation/metadata
  66 ;;; -    make sure that we do coherency checking in the exported
  67 ;;; -    functions.
  68 ;;; -    ...
  69 ;;; - reshapeData/reformat/reshapr a reformed version of the dataset (no
  70 ;;;           additional input).
  71 ;;; -         either overwriting or not, i.e. with or without copy.
  72 ;;; - check consistency of resulting data with metadata and related
  73 ;;;           data information.
  74 ;;; -
  75
  76 (defclass data-pointer ()
  77   ((store :initform nil
  78           :initarg :storage
  79           :accessor dataset
  80           :documentation "Data storage: typed as table, array,
  81                           relation, or pointer/reference to such.")
  82    (documentation-string :initform nil
  83                          :initarg :doc
  84                          :accessor doc-string
  85                          :documentation "uncomputable information
  86                                          about statistical-dataset
  87                                          instance.")
  88
  89    ;; the rest of this is metadata.  In particular, we should find a
  90    ;; more flexible, compact way to store this.
  91    (case-labels :initform nil
  92                 :initarg :case-labels
  93                 :accessor case-labels
  94                 :documentation "labels used for describing cases (doc
  95                                 metadata), possibly used for merging.")
  96    (var-labels :initform nil
  97                :initarg :var-labels
  98                :accessor var-labels
  99                :documentation "Variable names."))
 100   (:documentation "Standard Cases by Variables Statistical-Dataset,
 101                    i.e. an S data.frame."))
 102
 103
 104 (defgeneric get-variable-matrix (dataset-pointer list-of-variable-names)
 105   (:documentation "retrieves a matrix whose columns are the variable
 106   names in same order specified."))
 107
 108 (defgeneric get-variable-vector (dataset-pointer variable-name))
 109
 110 ;; statistical-dataset is the basic cases by variables framework.
 111 ;; Need to embed this within other structures which allow for
 112 ;; generalized relations.  Goal is to ensure that relations imply and
 113 ;; drive the potential for statistical relativeness such as
 114 ;; correlation, interference, and similar concepts.
 115 ;;
 116 ;; Actions on a statistical data structure.