Next: Introduction, Previous: (dir), Up: (dir) [Contents][Index]
This is the dense-arrays Reference Manual, version 0.2.0, generated automatically by Declt version 4.0 beta 2 "William Riker" on Tue Nov 15 05:35:51 2022 GMT+0.
Next: Systems, Previous: The dense-arrays Reference Manual, Up: The dense-arrays Reference Manual [Contents][Index]
[Last README update: 5th June 2022. To skip rant, click here.]
Table of Contents
Quick! Tell me the dimensions of
#2A((0 4 2 1 9 1 5 9 8 8 6 8 2 0 4 0 2 3 6 8 2 6 7 2 6 7 2 7 8 9 6 8 5 6 3 1 5
5 3 5 0 4 1 4 5 4 3 2 6 1 0 5 2 5 1 6 2 2 3 3 7 9 5 8 2 5 0 5 5 3 8 6 3 7
9 4 7 3 3 1 3 0 0 0 8 8 4 8 2 5 7 2 5 6 6 1 5 1 5 4 3 1 3 9 7 0 9 8)
(2 0 0 7 0 5 8 7 2 0 3 1 9 3 3 2 8 5 0 3 1 3 5 6 9 2 9 8 5 2 0 6 4 7 7 2 5
6 7 8 3 3 6 3 1 7 3 2 8 0 6 5 9 8 7 7 4 5 8 5 4 8 8 1 6 8 0 7 8 7 4 6 3 9
3 3 0 9 4 9 7 1 0 0 9 6 3 7 4 4 7 7 5 9 5 0 9 2 2 5 1 6 6 6 5 9 8 2)
(6 5 0 3 7 2 4 0 3 4 9 3 1 9 8 2 4 4 7 2 3 8 3 9 4 4 8 8 7 0 6 2 0 1 9 9 2
4 7 4 1 4 2 1 4 1 9 7 7 2 9 0 0 6 0 5 1 5 5 2 8 5 8 9 0 1 7 0 8 9 5 0 2 3
8 5 8 5 2 4 0 9 3 9 4 8 4 8 7 1 8 5 5 4 3 4 0 3 5 7 9 4 2 5 4 7 0 0)
(8 8 9 6 9 3 6 4 4 2 3 9 4 8 4 1 2 1 0 5 3 0 5 9 4 3 9 7 9 8 8 4 5 3 6 9 9
0 9 1 8 1 7 5 7 1 8 1 7 2 4 2 4 9 0 9 9 3 8 5 8 8 1 4 9 0 5 8 9 7 5 4 3 4
1 4 2 9 1 9 8 2 4 5 6 2 0 2 0 0 5 8 5 9 9 4 2 7 7 1 7 5 7 6 2 9 6 2)
(2 4 9 8 2 7 1 6 0 8 4 2 9 4 6 9 9 4 1 5 1 4 0 1 9 9 0 1 3 8 0 9 7 9 2 3 9
5 8 0 8 8 1 2 4 9 1 0 5 4 5 4 8 5 7 2 8 1 6 5 5 3 0 9 0 8 8 8 5 0 7 6 7 1
6 3 8 8 6 9 7 9 9 5 2 1 6 7 5 3 3 9 7 2 4 0 3 8 4 9 7 8 1 5 1 9 1 6)
(7 7 0 7 9 6 1 5 8 4 2 9 4 9 9 6 6 9 3 3 7 6 1 6 6 6 3 5 5 1 6 7 3 7 6 2 8
7 3 4 2 4 8 1 3 0 8 3 5 4 7 6 5 6 0 9 5 0 0 7 6 0 9 1 0 0 5 3 1 2 1 0 5 5
6 2 4 7 5 5 7 7 6 5 1 6 6 1 4 0 9 9 0 2 9 5 4 0 7 1 2 1 3 9 9 7 7 9))
and also the element-type!
CL-USER> (progn
(print (array-dimensions *))
(print (array-element-type *)))
(6 108)
T
T
Why do I need to do this every f*cking time? This can't be so annoying. I mean, why do I need to do all that to get something as basic as dimensions and element-type? Like magicl you mean?
#<MAGICL:MATRIX/SINGLE-FLOAT (6x108):
.
.
.
. >
See how pretty? Well not sarcastically. It is prettier than cl:array indeed, in that I can at least tell the element-type and dimensions at a glance, what more can you need?
(setq *print-length* 10)
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 6x108 T
(7 6 1 3 7 1 6 5 2 5 ...)
(8 7 6 9 7 5 5 1 0 0 ...)
(8 1 7 8 9 0 2 9 5 5 ...)
(0 5 8 6 2 6 9 2 5 3 ...)
(4 8 3 4 2 3 2 2 2 5 ...)
(7 8 2 5 5 3 2 8 6 1 ...)
{103F4CE453}>
Oh wait!
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 6x108 SINGLE-FLOAT
( 0.054 7.363 8.527 8.561 7.399 1.875
0.513 6.690 0.398 0.745 ...)
( 6.766 7.421 2.419 7.857 8.914 6.637
8.464 1.052 1.400 3.602 ...)
( 7.384 6.489 9.426 9.044 6.571 9.431
6.177 9.389 3.526 4.850 ...)
( 7.873 6.541 2.272 7.863 7.271 8.398
7.759 3.428 3.482 7.050 ...)
( 4.954 9.796 7.450 9.431 2.254 7.741
5.725 1.729 4.947 4.176 ...)
( 6.553 4.481 0.159 5.915 4.391 5.940
1.160 2.071 4.158 8.746 ...)
{103F81F7B3}>
No, wait!
So that first one is a 6x108 array with element-type T while the second has element-type single-float. And both have row-major layouts. Oh, and we even have their identity in case we want to check whether two objects are the same or just their elements are same!
We could have simply written a nice print-array function, couldn't we? Well, perhaps yes, but perhaps not! What do you mean?
(describe *)
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 6x108 SINGLE-FLOAT {103F81F7B3}>
[standard-object]
Slots with :INSTANCE allocation:
STORAGE = #(0.05400777 7.363088 8.526753 8.5613365 7.398881 1.8748772 0.5133271..
DIMENSIONS = (6 108)
ELEMENT-TYPE = SINGLE-FLOAT
RANK = 2
TOTAL-SIZE = 648
STRIDES = (108 1)
OFFSETS = (0 0)
LAYOUT = :ROW-MAJOR
ROOT-ARRAY = NIL
Storage, dimensions, element-type, rank, total-size, pretty obvious stuff.
Hmm, strides, offsets, layout, and root array: what... are these? Offsets, I think I know, it sounds like displaced-index-offset of cl:array, but wait, why is it a list and not a single number? Welcome to dense-arrays :)
But does that mean all the effort spent on magicl all these years will go to waste? Thankfully not! Turns out both magicl and dense-arrays - specifically standard-dense-array - are using cl:vector under the hood for storage:
(describe (magicl:rand '(6 108) :type 'single-float))
#<MAGICL:MATRIX/DOUBLE-FLOAT (6x108):..
[structure-object]
Slots with :INSTANCE allocation:
NROWS = 6
NCOLS = 108
SIZE = 648
LAYOUT = :COLUMN-MAJOR
STORAGE = #(0.23684204 0.7401872 0.8467122 0.17149377 0.5106092 0.10455426..
What that means is:
(ql:quickload "dense-arrays+magicl")
(rand 5 5 :type 'single-float)
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 5x5 SINGLE-FLOAT
( 0.058 0.553 0.918 0.653 0.080 )
( 0.283 0.289 0.843 0.236 0.333 )
( 0.434 0.701 0.826 0.339 0.306 )
( 0.275 0.021 0.774 0.248 0.841 )
( 0.426 0.913 0.662 0.058 0.395 )
{10438D6D13}>
(magicl-funcall #'magicl:svd *)
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 5x5 SINGLE-FLOAT
( -0.460 -0.292 0.740 0.354 0.174 )
( -0.401 0.179 0.112 -0.801 0.391 )
( -0.497 -0.221 -0.107 -0.209 -0.806 )
( -0.401 0.842 -0.068 0.348 -0.065 )
( -0.469 -0.352 -0.651 0.262 0.404 )
{10476B6C93}>
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 5x5 SINGLE-FLOAT
( 2.486 0.000e+00 0.000e+00 0.000e+00 0.000e+00)
( 0.000e+00 0.793 0.000e+00 0.000e+00 0.000e+00)
( 0.000e+00 0.000e+00 0.635 0.000e+00 0.000e+00)
( 0.000e+00 0.000e+00 0.000e+00 0.181 0.000e+00)
( 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.105 )
{10476B87F3}>
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 5x5 SINGLE-FLOAT
( -0.268 -0.465 -0.721 -0.278 -0.340 )
( 0.025 -0.717 0.150 -0.044 0.678 )
( -0.421 -0.360 0.320 0.660 -0.393 )
( -0.496 0.356 -0.444 0.402 0.519 )
( -0.710 0.115 0.398 -0.569 0.023 )
{10476B8D03}>
Welcome again to dense-arrays :D!
This system provides
dense-array-classdense-array with
dense-array, the storage object and associated meta-information could correspond to usual (cl:simple-array element-type 1) or static-vectors or cl-cudaDOC.org for more details about customization.*print-array* *print-length* *print-level* *print-lines* and is customizable via *array-element-print-format*; this could be improved and integrated further with builtins once someone wraps their head around The Lisp Pretty Printer.unupgraded-array and simple-unupgraded-array types that use (cl:simple-array * 1) for storage but do not upgrade the types; this can be helpful for better type checking*array-element-type* *array-element-type-alist* *dense-array-class* *array-layout*The multidimensional strides and offsets enable copy-free slicing and broadcasting.
CL-USER> (let ((a (make-array '(1000 1000))))
(time (loop for i below 1000 do (select:select a t i))))
Evaluation took:
0.159 seconds of real time
0.159541 seconds of total run time (0.159541 user, 0.000000 system)
100.63% CPU
42 lambdas converted
352,387,000 processor cycles
10,863,248 bytes consed
NIL
CL-USER> (let ((a (dense-arrays:make-array '(1000 1000))))
(time (loop for i below 1000 do (dense-arrays:aref a nil i))))
Evaluation took:
0.000 seconds of real time
0.000910 seconds of total run time (0.000878 user, 0.000032 system)
100.00% CPU
2,001,298 processor cycles
261,904 bytes consed
NIL
Limitations:
aref can be 5-6 times slower than cl:aref even after optimization; the work-around for this is do-arrays which can be up to 25% slower. See perf.org for example optimizations.cl:array is both a class and a specializing type-specifier; IIUC, non-builtins can only either be one of class or specializing type-specifier. However, this can now be rectified using extensible-compound-types, however one needs to also think about playing nice in the presence or absence of extensible-compound-types.(setf aref) and (setf row-major-aref) may need to be used using (funcall #'(setf aref) ...) since some implementations like SBCL "lose" the type information from the environment in an attempt to use once-only:layout has been provided recently, support for it can be buggy. Bug reports will be appreciated!dense-arrays: the super bare essentialsdense-arrays-plus-lite: some utilitiesdense-arrays+static-vectors: provides and exports a static-array type that is essentially a wrapper around static-vectorsdense-arrays+cuda: provides and export an array using cl-cudadense-arrays+magicl: provides helper functions magicl-funcall from-magicl-tensor as-magicl-tensor for interoperating between standard-dense-array and magicl:vector magicl:matrix magicl:tensordense-arrays-plus: more utilities as well as static-vectorsMinimalists would want to stick to the first four. The last one also introduces
shape as an alias for array-dimensionsint32 uint32 uint8 types as aliases for their common lisp counterpartspy4cl2:*array-type* to :dense-arrays when you want to use py4cl2 with dense-arrays.CL-USER> (uiop:define-package :dense-arrays-demo
(:mix :dense-arrays :cl))
#<PACKAGE "DENSE-ARRAYS-DEMO">
CL-USER> (in-package :dense-arrays-demo)
#<PACKAGE "DENSE-ARRAYS-DEMO">
CL-USER> (setq *print-length* 10) ; also intends to respect (*print-level* *print-lines* *print-array*)
10
A cleaner look
DENSE-ARRAYS-DEMO> (make-array '(2 3 4) :constructor #'+)
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 2x3x4 T
((0 1 2 3)
(1 2 3 4)
(2 3 4 5))
((1 2 3 4)
(2 3 4 5)
(3 4 5 6))
{100980B593}>
DENSE-ARRAYS-DEMO> (aref * 1 '(0 :step 2))
#<STANDARD-DENSE-ARRAY NIL 2x4 T
(1 2 3 4)
(3 4 5 6)
{1009810073}>
DENSE-ARRAYS-DEMO> (aref ** 1 '(-1 :step -2))
#<STANDARD-DENSE-ARRAY NIL 2x4 T
(3 4 5 6)
(1 2 3 4)
{1009811E73}>
DENSE-ARRAYS-DEMO> (make-array '(2 10))
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 2x10 T
(0 0 0 0 0 0 0 0 0 0)
(0 0 0 0 0 0 0 0 0 0)
{1009813D63}>
DENSE-ARRAYS-DEMO> (make-array '(2 100))
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 2x100 T
(0 0 0 0 0 0 0 0 0 0 ...)
(0 0 0 0 0 0 0 0 0 0 ...)
{10098286B3}>
DENSE-ARRAYS-DEMO> (describe (make-array '(2 10)))
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 2x10 T {100982C243}>
[standard-object]
Slots with :INSTANCE allocation:
STORAGE = #(0 0 0 0 0 0 0 0 0 0 ...)
DIMENSIONS = (2 10)
ELEMENT-TYPE = T
RANK = 2
TOTAL-SIZE = 20
STRIDES = (10 1)
OFFSETS = (0 0)
LAYOUT = :ROW-MAJOR
ROOT-ARRAY = NIL
; No value
DENSE-ARRAYS-DEMO> (defparameter a (make-array '(4 10) :constructor #'+))
A
DENSE-ARRAYS-DEMO> (print-array a "~2d")
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 4x10 T
( 0 1 2 3 4 5 6 7 8 9)
( 1 2 3 4 5 6 7 8 9 10)
( 2 3 4 5 6 7 8 9 10 11)
( 3 4 5 6 7 8 9 10 11 12)
{1009837073}>
NIL
Slicing facilities
DENSE-ARRAYS-DEMO> (aref a nil 1)
#<STANDARD-DENSE-ARRAY NIL 4 T
1
2
3
4
{100984D933}>
DENSE-ARRAYS-DEMO> (aref a nil -1)
#<STANDARD-DENSE-ARRAY NIL 4 T
9
10
11
12
{100984ECC3}>
DENSE-ARRAYS-DEMO> (aref a 1)
#<STANDARD-DENSE-ARRAY NIL 10 T
1
2
3
4
5
6
7
8
9
10
{100984FFA3}>
DENSE-ARRAYS-DEMO> (aref a '(1 :end 3) '(1 :end 3))
#<STANDARD-DENSE-ARRAY NIL 2x2 T
(2 3)
(3 4)
{10098622A3}>
DENSE-ARRAYS-DEMO> (defparameter b (aref a '(1 :end 3) '(1 :end 8 :step 2)))
B
DENSE-ARRAYS-DEMO> b
#<STANDARD-DENSE-ARRAY NIL 2x4 T
(2 4 6 8)
(3 5 7 9)
{1009863FF3}>
DENSE-ARRAYS-DEMO> (aref b (make-array '(2 4) :initial-contents '((0 1 0 0) (1 1 0 0))
:element-type 'bit))
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 3 T
4
3
5
{1009867F13}>
DENSE-ARRAYS-DEMO> (setf (aref b (make-array '(2 4)
:initial-contents '((0 1 0 0) (1 1 0 0))
:element-type 'bit))
0)
0
DENSE-ARRAYS-DEMO> b
#<STANDARD-DENSE-ARRAY NIL 2x4 T
(2 0 6 8)
(0 0 7 9)
{1009863FF3}>
DENSE-ARRAYS-DEMO> a
#<STANDARD-DENSE-ARRAY :ROW-MAJOR 4x10 T
(0 1 2 3 4 5 6 7 8 9)
(1 2 3 0 5 6 7 8 9 10)
(2 0 4 0 6 7 8 9 10 11)
(3 4 5 6 7 8 9 10 11 12)
{1009837073}>
Tests are also littered throughout out the system and may serve as examples, for instance plus/py4cl2.lisp.
digikar99/specialized-array-dispatch dist at https://github.com/digikar99/adhoc-polymorphic-functions/#getting-it-from-ultralisp$QUICKLISP_HOME/local-projects. (See ql:*local-project-directories*.)(ql:quickload "dense-arrays") - or dense-arrays-plus or dense-arrays-plus-lite(asdf:test-system "dense-arrays")- or dense-arrays-plus or dense-arrays-plus-liteFeel free to raise an issue!
Variable
Default Value: "~/DENSE-ARRAYS::PRETTY-PRINT-NUMBER/"
The format control string used to print the elements of dense-arrays:array.
It is possible to set this value to "~/USER-DEFINED-FUNCTION/" where USER-DEFINED-FUNCTION should accept at least four arguments.
Also see:
Variable
Default Unbound
If BOUND, this is the default value of the ELEMENT-TYPE or TYPE argument. Overrides *array-element-type-alist*. Is overriden by explicitly passing an ELEMENT-TYPE or TYPE argument.
Variable
Default Value: NIL
An ALIST mapping package to the default element-type used in that package. (Inspired from SWANK:READTABLE-ALIST) Overrides none. Is overriden by *array-element-type* when bound, or by explicitly passing an ELEMENT-TYPE or TYPE argument.
Variable
Default Value: :ROW-MAJOR
Specifies the default layout constructed by dense-arrays:make-array and constructor functions like asarray, zeros, ones, etc in the DENSE-ARRAYS-PLUS-LITE package.
Variable
Default Value: #<DENSE-ARRAYS:STANDARD-DENSE-ARRAY-CLASS DENSE-ARRAYS:STANDARD-DENSE-ARRAY>
Specifies the default value of CLASS in dense-arrays:make-array and other functions. (TODO: Specify these other functions.)
Polymorphic Function: (aref array &rest subscripts)
This is SETF-able.
((common-lisp:array common-lisp:array) &rest abstract-arrays::subscripts)A wrapper around CL:AREF.
Return the element of the array specified by the subscripts.
((array dense-array) &rest dense-arrays::subscripts)No documentation found.
Function: (aref* dense-array &rest subscripts)
Accessor function for DENSE-ARRAYS::DENSE-ARRAY with semantics intended to be similar to numpy's indexing semantics. See https://numpy.org/doc/stable/user/basics.indexing.html
Each element of subscripts can be
Each of the subscripts, START, END, STEP can also be negative integers, in which
case the last element along the axis is given the index -1, the second last is
given the index -2 and so on. Thus, (aref ... '(-1 :step -1)) can reverse a one
dimensional array.
Like, cl:aref or abstract-arrays:aref, returns the element corresponding to subscripts
if all the subscripts are integers and there as many subscripts
as the rank of the array.
If the number (aka length) of subscripts were less than the array's rank, or
if some of the subscripts were lists described above, then returns a VIEW
of the arrays. A VIEW is a window into the original array and thus
avoids copying the elements of the original array.
Examples illustrating the numpy-equivalent indexes:
a[::] (aref a nil)
a[::2] (aref a '(0 :step 2))
a[3, ::-1] (aref a 3 '(-1 :step -1))
a[3::, -1] (aref a '(3) -1)
The subscripts can also be integer or boolean arrays, denoting which elements
to select from each of the axes. But in this case the corresponding elements
of the array are copied over into a new array.
Type
A wrapper around STANDARD-DENSE-ARRAY with support for specifying ELEMENT-TYPE and DIMENSIONS or RANK. These specializers are the same like the CL:ARRAY compound type.
Function: (array-dimension array axis-number)
Return the length of dimension axis-number of array.
Polymorphic Function: (array-dimensions array)
((common-lisp:array common-lisp:array))No documentation found.
((common-lisp:array abstract-arrays:abstract-array))Returns a COPY of the dimensions of array. The copy may then be modified.
See narray-dimensions or equivalent of a copy is to be avoided, and destructive use is not intended.
Function: (array-displaced-to array)
Function: (array-displacement array)
Returns two values:
array is displaced along multiple axis.Polymorphic Function: (array-element-type array)
((common-lisp:array common-lisp:array))No documentation found.
((abstract-arrays:abstract-array abstract-arrays:abstract-array))No documentation found.
Function: (array-layout array)
Function: (array-offset array axis-number)
Return the length of offset corresponding to axis-number of array.
Function: (array-offsets array)
Polymorphic Function: (array-rank array)
((common-lisp:array common-lisp:array))No documentation found.
((abstract-arrays:abstract-array abstract-arrays:abstract-array))No documentation found.
Polymorphic Function: (array-storage array)
((abstract-arrays:abstract-array abstract-arrays:abstract-array))No documentation found.
((common-lisp:array common-lisp:array))No documentation found.
No documentation found for array-storage-allocator
No documentation found for array-storage-deallocator
Function: (array-stride array axis-number)
Return the length of stride corresponding to axis-number of array.
Function: (array-strides array)
Polymorphic Function: (array-total-size array)
((common-lisp:array common-lisp:array))No documentation found.
((abstract-arrays:abstract-array abstract-arrays:abstract-array))No documentation found.
Function: (array= array1 array2 &key (test (function equalp)))
Returns non-NIL if each element of array1 is equal to each corresponding
element of array2 using test, which should be a two-argument function that takes
the one element of the first array and the corresponding element of the second
and tests for their equality.
Function: (arrayp object)
Function: (as-cl-array array)
Function: (asarray array-like &key (out NIL outp) (type default-element-type)
(layout *array-layout*))
type can also be :AUTO
Function: (broadcast-array array broadcast-dimensions)
Function: (broadcast-arrays &rest arrays)
Returns two values. The first value is the list of broadcasted arrays if the second value is non-NIL.
Function: (broadcast-compatible-p &rest arrays)
Returns two values:
The broadcasting semantics are equivalent to numpy semantics. Two arrays are broadcast compatible, if
Thus, arrays with the following dimensions are broadcast-compatible:
Arrays with the following dimensions are not compatible:
See https://numpy.org/doc/stable/user/basics.broadcasting.html for an elaborate discussion.
Function: (copy-array array &key (layout (array-layout array)))
Returns a copy of array. Creates a completely new array even if array
is a VIEW (see ARRAY-VIEW-P).
No documentation found for copy-dense-array
Macro: (define-array-class name &body (direct-slots &rest slot-options))
Defines name as a CLASS with DIRECT-SUPERCLASS ABSTRACT-ARRAY and metaclass
as ABSTRACT-ARRAY-CLASS. Also defines the appropriate order using direct-slots.
Function: (dense-array-type-class array-type &optional env)
Macro: (do-arrays rank/bindings &body body)
Traverses the arrays in row-major order.
If the first argument rank/bindings is of type SIZE, it'd be treated as the rank
of the arrays. Then, the BINDINGS are assumed to be the first element of the body.
Otherwise, the first argument is treated as if they are BINDINGS. Each BINDING is of the form (ELT-VAR array &OPTIONAL (ELEMENT-TYPE *) &KEY (CLASS-NAME *dense-array-class*)) Here, only array is evaluated.
Examples
(let ((a (make-array '(2 3)))
(b (make-array '(2 3))))
(do-arrays ((c a t)
(d b t))
(print (list c d))))
(let ((a (make-array '(2 3)))
(b (make-array '(2 3))))
(do-arrays 2 ((c a t) ; The 2 indicates the rank of the arrays
(d b t))
(print (list c d))))
Either of the two cases might be faster depending on the number of dimensions.
Function: (eye per-axis-size &key (rank 2) (type default-element-type))
Function: (full &rest args)
LAMBDA-LIST: (SHAPE &KEY (TYPE DEFAULT-ELEMENT-TYPE) (LAYOUT *array-layout*) VALUE)
Function: (full-like array-like value)
Macro: (macro-map-array result-array function &rest arrays)
Function: (make-array dimensions &rest args &key
(element-type default-element-type)
(initial-element NIL initial-element-p)
(initial-contents NIL initial-contents-p)
(constructor NIL constructor-p) (strides NIL strides-p)
(adjustable NIL adjustable-p) (fill-pointer NIL fill-pointer-p)
(class *dense-array-class*) (layout *array-layout*)
(displaced-to NIL displaced-to-p) (offsets NIL offsets-p)
(displaced-index-offset 0 displaced-index-offset-p))
Like CL:MAKE-ARRAY but returns a DENSE-ARRAYS::DENSE-ARRAY instead of cl:array. Additionally takes
layout argument which can be one of (:ROW-MAJOR :COLUMN-MAJOR NIL)
class argument which should be a class designator denoting the class to which the
constructed dense array will belong to
constructor if supplied should be a function that takes as many arguments
as the number of dimensions aka rank of the array, and return the element that
should correspond to the position indicated by the arguments of the function.
For example:
(make-array '(2 3) :constructor (lambda (&rest indexes) (cons 'indexes indexes))) ;=> #<standard-dense-array :ROW-MAJOR 2x3 T ((INDEXES 0 0) (INDEXES 0 1) (INDEXES 0 2)) ((INDEXES 1 0) (INDEXES 1 1) (INDEXES 1 2)) {10194A2FE3}>
Polymorphic Function: (narray-dimensions array)
((common-lisp:array abstract-arrays:abstract-array))Returns the dimensions of the array. The consequences are undefined if the
returned dimensions are modified. Use array-dimensions if destructive usage of
the returned list is intended.
Function: (ones &rest args)
LAMBDA-LIST: (SHAPE &KEY (TYPE DEFAULT-ELEMENT-TYPE) (LAYOUT *array-layout*))
Function: (ones-like array-like)
Function: (print-array array &optional array-element-print-format &key level
length (stream NIL streamp))
Prints array as if by CL:PRINT.
Format recipes: http://www.gigamonkeys.com/book/a-few-format-recipes.html.
Function: (rand &rest args)
LAMBDA-LIST: (SHAPE &KEY (TYPE DEFAULT-ELEMENT-TYPE) (LAYOUT *array-layout*) (MIN (COERCE 0 TYPE)) (MAX (COERCE 1 TYPE)))
Function: (rand-like array-like)
Function: (reshape array-like new-shape &key (view NIL viewp)
(layout NIL layoutp))
view argument is considered only if array-like is a SIMPLE-DENSE-ARRAY.
If array-like is a SIMPLE-DENSE-ARRAY, it is guaranteed that when view is supplied,
array-like is createdarray-like is not a SIMPLE-DENSE-ARRAY,
then a new array is created. In the future, an attempt may be made to avoid
creating the new array and instead return a view instead.Polymorphic Function: (row-major-aref array index)
Return the element of array corresponding to the row-major index.
This is SETFable.
((common-lisp:array common-lisp:array) (abstract-arrays::index t))No documentation found.
((array dense-array) (dense-arrays::index t))No documentation found.
Type: (SIMPLE-ARRAY &OPTIONAL (ABSTRACT-ARRAYS::ELEMENT-TYPE '*)
(ABSTRACT-ARRAYS::DIM/RANK '*))
A wrapper around (AND STANDARD-DENSE-ARRAY SIMPLE-DENSE-ARRAY) with support for specifying ELEMENT-TYPE and DIMENSIONS or RANK. These specializers are the same like the CL:ARRAY compound type.
Type: (SIMPLE-UNUPGRADED-ARRAY &OPTIONAL (ABSTRACT-ARRAYS::ELEMENT-TYPE '*)
(ABSTRACT-ARRAYS::DIM/RANK '*))
A wrapper around (AND UNUPGRADED-DENSE-ARRAY SIMPLE-DENSE-ARRAY) with support for specifying ELEMENT-TYPE and DIMENSIONS or RANK. These specializers are the same like the CL:ARRAY compound type.
Class
Class
Generic Function: (storage-accessor class)
Returns a SYMBOL that is fbound to an accessor function that takes (STORAGE INDEX) as arguments and returns the element at INDEX in STORAGE. The function is an accessor function in the sense that SYMBOL should also be associated with (SETF SYMBOL) function that takes (NEW-VALUE STORAGE INDEX) as arguments and sets the STORAGE element at INDEX to NEW-VALUE. This function is primarily used inside aref, row-major-aref and do-arrays, and their SETF counterparts. See src/protocol.lisp and plus/cl-cuda.lisp for reference.
Generic Function: (storage-allocator class)
Returns a symbol fbound to a function with signature (SIZE &KEY ELEMENT-TYPE INITIAL-ELEMENT) that allocates a VECTOR of length SIZE of ELEMENT-TYPE with each element as INITIAL-ELEMENT for use as a STORAGE-VECTOR for the ABSTRACT-ARRAY.
Generic Function: (storage-deallocator class)
Returns either NIL or a symbol fbound to a function to be called to delete the STORAGE when the ABSTRACT-ARRAY goes out of scope. This function should take only the STORAGE object as its argument. Internally, this function plays a role in the finalizer of the garbage collection using TRIVIAL-GARBAGE. See plus/static-vectors.lisp and the dense-arrays:make-array function for reference.
Generic Function: (storage-element-type-upgrader class)
Equivalent to the CL:UPGRADED-ARRAY-ELEMENT-TYPE, this returns a function
that takes a single argument element-type as input and returns the upgraded
array element type for the array class given by class used for STORAGE.
The upgraded array element type is then stored in the dense-array object and
used for other tasks downstream.
See plus/cl-cuda.lisp and the dense-arrays:make-array function for reference.
Generic Function: (storage-type-inferrer-from-array-type class)
This should return a function that takes as input the ARRAY-TYPE and returns the possibly specialized type of storage that the corresponding array object will have. This is primarily used for optimization purposes inside dense-arrays:do-arrays and the compiler macros of dense-arrays:aref dense-arrays:row-major-aref and SETF counterparts. See src/protocol.lisp, plus/cl-cuda.lisp, src/do-arrays.lisp and optim/aref.lisp for reference.
Function: (transpose array-like &key axes)
Type: (UNUPGRADED-ARRAY &OPTIONAL (ABSTRACT-ARRAYS::ELEMENT-TYPE '*)
(ABSTRACT-ARRAYS::DIM/RANK '*))
A wrapper around UNUPGRADED-DENSE-ARRAY with support for specifying ELEMENT-TYPE and DIMENSIONS or RANK. These specializers are the same like the CL:ARRAY compound type.
Class
Function: (zeros &rest args)
LAMBDA-LIST: (SHAPE &KEY (TYPE DEFAULT-ELEMENT-TYPE) (LAYOUT *array-layout*))
Function: (zeros-like array-like)
Next: Modules, Previous: Introduction, Up: The dense-arrays Reference Manual [Contents][Index]
The main system appears first, followed by any subsystem dependency.
Numpy like array objects for Common Lisp
Shubhamkar B. Ayare
MIT
0.2.0
Next: Files, Previous: Systems, Up: The dense-arrays Reference Manual [Contents][Index]
Modules are listed depth-first from the system components tree.
Next: dense-arrays/optim, Previous: Modules, Up: Modules [Contents][Index]
dense-arrays (system).
Previous: dense-arrays/src, Up: Modules [Contents][Index]
src (module).
dense-arrays (system).
Next: Packages, Previous: Modules, Up: The dense-arrays Reference Manual [Contents][Index]
Files are sorted by type and then listed depth-first from the systems components trees.
Next: dense-arrays/src/package.lisp, Previous: Lisp, Up: Lisp [Contents][Index]
dense-arrays (system).
Next: dense-arrays/src/protocol.lisp, Previous: dense-arrays/dense-arrays.asd, Up: Lisp [Contents][Index]
src (module).
Next: dense-arrays/src/types.lisp, Previous: dense-arrays/src/package.lisp, Up: Lisp [Contents][Index]
package.lisp (file).
src (module).
Next: dense-arrays/src/dense-arrays.lisp, Previous: dense-arrays/src/protocol.lisp, Up: Lisp [Contents][Index]
protocol.lisp (file).
src (module).
Next: dense-arrays/src/broadcast.lisp, Previous: dense-arrays/src/types.lisp, Up: Lisp [Contents][Index]
types.lisp (file).
src (module).
Next: dense-arrays/src/do-arrays.lisp, Previous: dense-arrays/src/dense-arrays.lisp, Up: Lisp [Contents][Index]
dense-arrays.lisp (file).
src (module).
%broadcast-compatible-p (function).
Next: dense-arrays/src/argwhere.lisp, Previous: dense-arrays/src/broadcast.lisp, Up: Lisp [Contents][Index]
broadcast.lisp (file).
src (module).
do-arrays (macro).
Next: dense-arrays/src/aref.lisp, Previous: dense-arrays/src/do-arrays.lisp, Up: Lisp [Contents][Index]
do-arrays.lisp (file).
src (module).
Next: dense-arrays/src/copy.lisp, Previous: dense-arrays/src/argwhere.lisp, Up: Lisp [Contents][Index]
argwhere.lisp (file).
src (module).
Next: dense-arrays/src/unupgraded.lisp, Previous: dense-arrays/src/aref.lisp, Up: Lisp [Contents][Index]
aref.lisp (file).
src (module).
copy-array (function).
Next: dense-arrays/optim/misc.lisp, Previous: dense-arrays/src/copy.lisp, Up: Lisp [Contents][Index]
copy.lisp (file).
src (module).
unupgraded-dense-array-class (class).
Next: dense-arrays/optim/aref.lisp, Previous: dense-arrays/src/unupgraded.lisp, Up: Lisp [Contents][Index]
optim (module).
dense-array-type-class (function).
Next: dense-arrays/optim/unupgraded.lisp, Previous: dense-arrays/optim/misc.lisp, Up: Lisp [Contents][Index]
misc.lisp (file).
optim (module).
array= (compiler macro).
Next: Definitions, Previous: Files, Up: The dense-arrays Reference Manual [Contents][Index]
Packages are listed by definition order.
Next: Indexes, Previous: Packages, Up: The dense-arrays Reference Manual [Contents][Index]
Definitions are sorted by export status, category, package, and then by lexicographic order.
Next: Internals, Previous: Definitions, Up: Definitions [Contents][Index]
Next: Macros, Previous: Public Interface, Up: Public Interface [Contents][Index]
The format control string used to print the elements of DENSE-ARRAYS:ARRAY.
It is possible to set this value to "~/USER-DEFINED-FUNCTION/" where
USER-DEFINED-FUNCTION should accept at least four arguments.
Also see:
- https://en.wikipedia.org/wiki/Format_(Common_Lisp)
- http://www.gigamonkeys.com/book/a-few-format-recipes.html
If BOUND, this is the default value of the ELEMENT-TYPE or TYPE argument.
Overrides *ARRAY-ELEMENT-TYPE-ALIST*.
Is overriden by explicitly passing an ELEMENT-TYPE or TYPE argument.
An ALIST mapping package to the default element-type used in that package.
(Inspired from SWANK:*READTABLE-ALIST*)
Overrides none.
Is overriden by *ARRAY-ELEMENT-TYPE* when bound, or by explicitly passing an
ELEMENT-TYPE or TYPE argument.
Specifies the default layout constructed by DENSE-ARRAYS:MAKE-ARRAY and constructor functions like ASARRAY, ZEROS, ONES, etc in the DENSE-ARRAYS-PLUS-LITE package.
Specifies the default value of CLASS in DENSE-ARRAYS:MAKE-ARRAY and other functions. (TODO: Specify these other functions.)
Next: Compiler macros, Previous: Special variables, Up: Public Interface [Contents][Index]
Traverses the arrays in row-major order.
If the first argument RANK/BINDINGS is of type SIZE, it’d be treated as the rank of the arrays. Then, the BINDINGS are assumed to be the first element of the BODY.
Otherwise, the first argument is treated as if they are BINDINGS.
Each BINDING is of the form
(ELT-VAR ARRAY &OPTIONAL (ELEMENT-TYPE *) &KEY (CLASS-NAME *DENSE-ARRAY-CLASS*))
Here, only ARRAY is evaluated.
Examples
(let ((a (make-array ’(2 3)))
(b (make-array ’(2 3))))
(do-arrays ((c a t)
(d b t))
(print (list c d))))
(let ((a (make-array ’(2 3)))
(b (make-array ’(2 3))))
(do-arrays 2 ((c a t) ; The 2 indicates the rank of the arrays
(d b t))
(print (list c d))))
Either of the two cases might be faster depending on the number of dimensions.
Next: Ordinary functions, Previous: Macros, Up: Public Interface [Contents][Index]
Next: Generic functions, Previous: Compiler macros, Up: Public Interface [Contents][Index]
Accessor function for DENSE-ARRAYS::DENSE-ARRAY with semantics
intended to be similar to numpy’s indexing semantics.
See https://numpy.org/doc/stable/user/basics.indexing.html
Each element of SUBSCRIPTS can be
- either an integer denoting the position within the axis which is to be indexed
- or a list of the form (&OPTIONAL START &KEY END STEP) with each of START END
STEP being integers if supplied. START denotes the start position within the
axis, END denotes the ending position within the axis, STEP denotes at what
distance within the axis the next element should come after the previous,
starting from START
Each of the SUBSCRIPTS, START, END, STEP can also be negative integers, in which
case the last element along the axis is given the index -1, the second last is
given the index -2 and so on. Thus, ‘(aref ... ’(-1 :step -1))‘ can reverse a one
dimensional array.
Like, CL:AREF or ABSTRACT-ARRAYS:AREF, returns the element corresponding to SUBSCRIPTS
if all the subscripts are integers and there as many subscripts
as the rank of the array.
If the number (aka length) of SUBSCRIPTS were less than the array’s rank, or
if some of the SUBSCRIPTS were lists described above, then returns a VIEW
of the arrays. A VIEW is a window into the original array and thus
avoids copying the elements of the original array.
Examples illustrating the numpy-equivalent indexes:
a[::] (aref a nil)
a[::2] (aref a ’(0 :step 2))
a[3, ::-1] (aref a 3 ’(-1 :step -1))
a[3::, -1] (aref a ’(3) -1)
The SUBSCRIPTS can also be integer or boolean arrays, denoting which elements
to select from each of the axes. But in this case the corresponding elements
of the array are copied over into a new array.
Returns two values:
- ARRAY-STORAGE
- and OFFSET along first axis
Consequences are undefined if ARRAY is displaced along multiple axis.
Return the length of offset corresponding to AXIS-NUMBER of ARRAY.
Return the length of stride corresponding to AXIS-NUMBER of ARRAY.
Returns non-NIL if each element of ARRAY1 is equal to each corresponding element of ARRAY2 using TEST, which should be a two-argument function that takes the one element of the first array and the corresponding element of the second and tests for their equality.
Returns two values. The first value is the list of broadcasted arrays if the second value is non-NIL.
Returns two values:
- The first value is a generalized boolean indicating whether the arrays can be broadcasted.
- The second value is the dimension of the array resulting from the broadcast.
The broadcasting semantics are equivalent to numpy semantics. Two arrays are broadcast
compatible, if
- they have the same dimensions, or
- if, for the dimensions they differ, one of the dimension is of length 1, or
- the dimensions of the lower-ranked array matches the rightmost dimensions
of the higher-ranked array
Thus, arrays with the following dimensions are broadcast-compatible:
- (3) (3)
- (3 1) (3 3)
- (3 3) (1 3)
- (3 3) (3)
Arrays with the following dimensions are not compatible:
- (3 1) (3)
See https://numpy.org/doc/stable/user/basics.broadcasting.html for an elaborate discussion.
Returns a copy of ARRAY. Creates a completely new array even if ARRAY is a VIEW (see ARRAY-VIEW-P).
Like CL:MAKE-ARRAY but returns a DENSE-ARRAYS::DENSE-ARRAY instead of CL:ARRAY.
Additionally takes
- LAYOUT argument which can be one of (:ROW-MAJOR :COLUMN-MAJOR NIL)
- CLASS argument which should be a class designator denoting the class to which the
constructed dense array will belong to
- CONSTRUCTOR if supplied should be a function that takes as many arguments
as the number of dimensions aka rank of the array, and return the element that
should correspond to the position indicated by the arguments of the function.
For example:
(make-array ’(2 3) :constructor (lambda (&rest indexes) (cons ’indexes indexes)))
;=> #<STANDARD-DENSE-ARRAY :ROW-MAJOR 2x3 T
((INDEXES 0 0) (INDEXES 0 1) (INDEXES 0 2))
((INDEXES 1 0) (INDEXES 1 1) (INDEXES 1 2))
{10194A2FE3}>
Prints ARRAY as if by CL:PRINT.
Format recipes: http://www.gigamonkeys.com/book/a-few-format-recipes.html.
Next: Standalone methods, Previous: Ordinary functions, Up: Public Interface [Contents][Index]
Returns a SYMBOL that is fbound to an accessor function that takes
(STORAGE INDEX) as arguments and returns the element at INDEX in STORAGE.
The function is an accessor function in the sense that SYMBOL should also be
associated with (SETF SYMBOL) function that takes (NEW-VALUE STORAGE INDEX) as
arguments and sets the STORAGE element at INDEX to NEW-VALUE.
This function is primarily used inside AREF, ROW-MAJOR-AREF and DO-ARRAYS,
and their SETF counterparts.
See src/protocol.lisp and plus/cl-cuda.lisp for reference.
Returns a symbol fbound to a function with signature
(SIZE &KEY ELEMENT-TYPE INITIAL-ELEMENT)
that allocates a VECTOR of length SIZE of ELEMENT-TYPE with each element as
INITIAL-ELEMENT for use as a STORAGE-VECTOR for the ABSTRACT-ARRAY.
Returns either NIL or a symbol fbound to a function to be called
to delete the STORAGE when the ABSTRACT-ARRAY goes out of scope. This function should
take only the STORAGE object as its argument.
Internally, this function plays a role in the finalizer of the garbage collection
using TRIVIAL-GARBAGE.
See plus/static-vectors.lisp and the DENSE-ARRAYS:MAKE-ARRAY function for reference.
Equivalent to the CL:UPGRADED-ARRAY-ELEMENT-TYPE, this returns a function
that takes a single argument element-type as input and returns the upgraded
array element type for the array class given by CLASS used for STORAGE.
The upgraded array element type is then stored in the dense-array object and
used for other tasks downstream.
See plus/cl-cuda.lisp and the DENSE-ARRAYS:MAKE-ARRAY function for reference.
This should return a function that takes as input the ARRAY-TYPE and returns
the possibly specialized type of storage that the corresponding array object
will have. This is primarily used for optimization purposes inside DENSE-ARRAYS:DO-ARRAYS
and the compiler macros of DENSE-ARRAYS:AREF DENSE-ARRAYS:ROW-MAJOR-AREF and SETF
counterparts.
See src/protocol.lisp, plus/cl-cuda.lisp, src/do-arrays.lisp and optim/aref.lisp
for reference.
Next: Classes, Previous: Generic functions, Up: Public Interface [Contents][Index]
Next: Types, Previous: Standalone methods, Up: Public Interface [Contents][Index]
Previous: Classes, Up: Public Interface [Contents][Index]
Previous: Public Interface, Up: Definitions [Contents][Index]
Next: Ordinary functions, Previous: Symbol macros, Up: Internals [Contents][Index]
(map-collect ‘(c ,%1) ’(1 2 3)) ;=> ((C 1) (C 2) (C 3))
Next: Generic functions, Previous: Macros, Up: Internals [Contents][Index]
Returns a copy of the subscripted array.
Returns a VIEW of the subscripted ARRAY (without copying the contents)
Returns two values:
The first value is a generalized boolean indicating whether the two dimensions are broadcast compatible.
The second value is the dimensions of the array resulting from the broadcast.
Returns the storage-vector underlying the ARRAY. This is equivalent to ARRAY-DISPLACED-TO.
Returns T if the ARRAY is a VIEW, otherwise returns NIL. A VIEW does not
have a known layout, and is useful as a window into certain elements of the larger array.
A VIEW provides a way to obtain sub-arrays or reshapes or transposes out of ARRAY
without copying.
Next: Conditions, Previous: Ordinary functions, Up: Internals [Contents][Index]
Next: Classes, Previous: Generic functions, Up: Internals [Contents][Index]
:class
(quote (quote dense-arrays::backend))
:slot
optimization-failure-note.
:binding-form
:class
(quote (quote abstract-arrays::element-type))
:slot
error.
:index
condition.
:axis-number
:class
(quote (quote abstract-arrays::rank))
:slot
optimization-failure-note.
Next: Types, Previous: Conditions, Up: Internals [Contents][Index]
- DIMENSIONS is a list of dimensions.
- STRIDES is a list of strides along each dimension.
- OFFSETS is a list of offsets along each dimension.
abstract-array.
(error "strides must be supplied during dense-array initialization")
:strides
(error "offsets must be supplied during dense-array initialization")
:offsets
(member nil :row-major :column-major)
(error "layout must be supplied during dense-array initialization")
:layout
(error "root-array must be supplied during dense-array initialization")
:root-array
abstract-array-class.
Previous: Definitions, Up: The dense-arrays Reference Manual [Contents][Index]
| Jump to: | %
(
1
A B C D E F G I L M N P S T |
|---|
| Jump to: | %
(
1
A B C D E F G I L M N P S T |
|---|
Next: Data types, Previous: Functions, Up: Indexes [Contents][Index]
| Jump to: | *
A B D F I L O P R S |
|---|
| Jump to: | *
A B D F I L O P R S |
|---|
| Jump to: | %
A B C D E F I M O P R S T U |
|---|
| Jump to: | %
A B C D E F I M O P R S T U |
|---|