phobos/std/experimental/ndslice/package.d
Sebastian Wilzbach 2dfbc51f17 Standardize whitespace after imports
Unified with:

sed -E "s/import\s*([^ ]+)\s*:\s*(.*(,|;))/import \1 : \2/" -i **/*.d
2016-05-29 22:09:56 +02:00

590 lines
17 KiB
D
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/+
## Guide for Slice/BLAS contributors
1. Make sure functions are
a. inlined(!),
b. `@nogc`,
c. `nothrow`,
d. `pure`.
For this reason, it is preferable to use _simple_ `assert`s with messages
that can be computed at compile time.
The goals are:
1. to reduce executable size for _any_ compilation mode
2. to reduce template bloat in object files
3. to reduce compilation time
4. to allow users to write extern C bindings for code libraries on `Slice` type.
2. `std.format`, `std.string`, and `std.conv` should not be used in error
message formatting.`"Use" ~ Concatenation.stringof`.
3. `mixin template`s may be used for pretty error message formatting.
4. `Exception`s/`enforce`s should no be used to check indexes and lengths.
Exceptions are only allowed for algorithms where validation of input data is
too complicated for the user. `reshape` function is a good example of a case
where Exceptions are required.
If a function might throw an exception, an example with exception handing should be added.
5. For simple checks like matrix transposition, compile time flags should not be used.
It is much better to opt for runtime matrix transposition.
Furthermore, Slice type provides runtime matrix transposition out of the box.
6. _Fortran_VS_C_ flags should not be used. They are about notation,
but not about the algorithm itself. For math world users,
a corresponding code example might be included in the documentation.
`transposed` / `everted` can be used in cache-friendly codes.
7. Compile time evaluation should not be used to produce dummy types like `IdentityMatrix`.
8. Memory allocation and algorithm logic should be separated whenever possible.
9. CTFE unittests should be added to new functions.
+/
/**
$(H1 Multidimensional Random Access Ranges)
The package provides a multidimensional array implementation.
It would be well suited to creating machine learning and image
processing algorithms, but should also be general enough for use anywhere with
homogeneously-typed multidimensional data.
In addition, it includes various functions for iteration, accessing, and manipulation.
Quick_Start:
$(SUBREF slice, sliced) is a function designed to create
a multidimensional view over a range.
Multidimensional view is presented by $(SUBREF slice, Slice) type.
------
auto matrix = slice!double(3, 4);
matrix[] = 0;
matrix.diagonal[] = 1;
auto row = matrix[2];
row[3] = 6;
assert(matrix[2, 3] == 6); // D & C index order
assert(matrix(3, 2) == 6); // Math & Fortran index order
------
Note:
In many examples $(LINK2 std_experimental_ndslice_selection.html#iotaSlice, iotaSlice) is used
instead of a regular array, which makes it
possible to carry out tests without memory allocation.
$(SCRIPT inhibitQuickIndex = 1;)
$(DIVC quickindex,
$(BOOKTABLE ,
$(TR $(TH Category) $(TH Submodule) $(TH Declarations)
)
$(TR $(TDNW Basic Level
$(BR) $(SMALL $(SUBREF slice, Slice), its properties, operator overloading))
$(TDNW $(SUBMODULE slice))
$(TD
$(SUBREF slice, sliced)
$(SUBREF slice, Slice)
$(SUBREF slice, slice)
$(SUBREF slice, makeSlice)
$(SUBREF slice, ndarray)
$(SUBREF slice, makeNdarray)
$(SUBREF slice, shape)
$(SUBREF slice, Slice)
$(SUBREF slice, assumeSameStructure)
$(SUBREF slice, ReplaceArrayWithPointer)
$(SUBREF slice, DeepElementType)
$(SUBREF slice, SliceException)
)
)
$(TR $(TDNW Medium Level
$(BR) $(SMALL Various iteration operators))
$(TDNW $(SUBMODULE iteration))
$(TD
$(SUBREF iteration, transposed)
$(SUBREF iteration, strided)
$(SUBREF iteration, reversed)
$(SUBREF iteration, rotated)
$(SUBREF iteration, everted)
$(SUBREF iteration, swapped)
$(SUBREF iteration, allReversed)
$(SUBREF iteration, dropToHypercube) and other `drop` primitives
)
)
$(TR $(TDNW Advanced Level $(BR)
$(SMALL Abstract operators for loop free programming
$(BR) Take `movingWindowByChannel` as an example))
$(TDNW $(SUBMODULE selection))
$(TD
$(SUBREF selection, blocks)
$(SUBREF selection, windows)
$(SUBREF selection, diagonal)
$(SUBREF selection, reshape)
$(SUBREF selection, byElement)
$(SUBREF selection, byElementInStandardSimplex)
$(SUBREF selection, indexSlice)
$(SUBREF selection, iotaSlice)
$(SUBREF selection, pack)
$(SUBREF selection, evertPack)
$(SUBREF selection, unpack)
$(SUBREF selection, ReshapeException)
)
)
))
$(H2 Example: Image Processing)
A median filter is implemented as an example. The function
`movingWindowByChannel` can also be used with other filters that use a sliding
window as the argument, in particular with convolution matrices such as the
$(LINK2 https://en.wikipedia.org/wiki/Sobel_operator, Sobel operator).
`movingWindowByChannel` iterates over an image in sliding window mode.
Each window is transferred to a `filter`, which calculates the value of the
pixel that corresponds to the given window.
This function does not calculate border cases in which a window overlaps
the image partially. However, the function can still be used to carry out such
calculations. That can be done by creating an amplified image, with the edges
reflected from the original image, and then applying the given function to the
new file.
Note: You can find the example at
$(LINK2 https://github.com/libmir/mir/blob/master/examples/median_filter.d, GitHub).
-------
/++
Params:
filter = unary function. Dimension window 2D is the argument.
image = image dimensions `(h, w, c)`,
where с is the number of channels in the image
nr = number of rows in the window
nс = number of columns in the window
Returns:
image dimensions `(h - nr + 1, w - nc + 1, c)`,
where с is the number of channels in the image.
Dense data layout is guaranteed.
+/
Slice!(3, C*) movingWindowByChannel(alias filter, C)
(Slice!(3, C*) image, size_t nr, size_t nc)
{
import std.algorithm.iteration : map;
import std.array : array;
// 0. 3D
// The last dimension represents the color channel.
auto wnds = image
// 1. 2D composed of 1D
// Packs the last dimension.
.pack!1
// 2. 2D composed of 2D composed of 1D
// Splits image into overlapping windows.
.windows(nr, nc)
// 3. 5D
// Unpacks the windows.
.unpack
// 4. 5D
// Brings the color channel dimension to the third position.
.transposed!(0, 1, 4)
// 5. 3D Composed of 2D
// Packs the last two dimensions.
.pack!2;
return wnds
// 6. Range composed of 2D
// Gathers all windows in the range.
.byElement
// 7. Range composed of pixels
// 2D to pixel lazy conversion.
.map!filter
// 8. `C[]`
// The only memory allocation in this function.
.array
// 9. 3D
// Returns slice with corresponding shape.
.sliced(wnds.shape);
}
-------
A function that calculates the value of iterator median is also necessary.
-------
/++
Params:
r = input range
buf = buffer with length no less than the number of elements in `r`
Returns:
median value over the range `r`
+/
T median(Range, T)(Range r, T[] buf)
{
import std.algorithm.sorting : topN;
size_t n;
foreach (e; r)
buf[n++] = e;
auto m = n >> 1;
buf[0 .. n].topN(m);
return buf[m];
}
-------
The `main` function:
-------
void main(string[] args)
{
import std.conv : to;
import std.getopt : getopt, defaultGetoptPrinter;
import std.path : stripExtension;
uint nr, nc, def = 3;
auto helpInformation = args.getopt(
"nr", "number of rows in window, default value is " ~ def.to!string, &nr,
"nc", "number of columns in window, default value is equal to nr", &nc);
if (helpInformation.helpWanted)
{
defaultGetoptPrinter(
"Usage: median-filter [<options...>] [<file_names...>]\noptions:",
helpInformation.options);
return;
}
if (!nr) nr = def;
if (!nc) nc = nr;
auto buf = new ubyte[nr * nc];
foreach (name; args[1 .. $])
{
import imageformats; // can be found at code.dlang.org
IFImage image = read_image(name);
auto ret = image.pixels
.sliced(cast(size_t)image.h, cast(size_t)image.w, cast(size_t)image.c)
.movingWindowByChannel
!(window => median(window.byElement, buf))
(nr, nc);
write_image(
name.stripExtension ~ "_filtered.png",
ret.length!1,
ret.length!0,
(&ret[0, 0, 0])[0 .. ret.elementsCount]);
}
}
-------
This program works both with color and grayscale images.
-------
$ median-filter --help
Usage: median-filter [<options...>] [<file_names...>]
options:
--nr number of rows in window, default value is 3
--nc number of columns in window default value equals to nr
-h --help This help information.
-------
$(H2 Compared with `numpy.ndarray`)
numpy is undoubtedly one of the most effective software packages that has
facilitated the work of many engineers and scientists. However, due to the
specifics of implementation of Python, a programmer who wishes to use the
functions not represented in numpy may find that the built-in functions
implemented specifically for numpy are not enough, and their Python
implementations work at a very low speed. Extending numpy can be done, but
is somewhat laborious as even the most basic numpy functions that refer
directly to `ndarray` data must be implemented in C for reasonable performance.
At the same time, while working with `ndslice`, an engineer has access to the
whole set of standard D library, so the functions he creates will be as
efficient as if they were written in C.
License: $(WEB www.boost.org/LICENSE_1_0.txt, Boost License 1.0).
Authors: Ilya Yaroshenko
Acknowledgements: John Loughran Colvin
Source: $(PHOBOSSRC std/_experimental/_ndslice/_package.d)
Macros:
SUBMODULE = $(MREF std, experimental, ndslice, $1)
SUBREF = $(REF_ALTTEXT $(TT $2), $2, std,experimental, ndslice, $1)$(NBSP)
T2=$(TR $(TDNW $(LREF $1)) $(TD $+))
T4=$(TR $(TDNW $(LREF $1)) $(TD $2) $(TD $3) $(TD $4))
*/
module std.experimental.ndslice;
public import std.experimental.ndslice.slice;
public import std.experimental.ndslice.iteration;
public import std.experimental.ndslice.selection;
unittest
{
auto matrix = new double[12].sliced(3, 4);
matrix[] = 0;
matrix.diagonal[] = 1;
auto row = matrix[2];
row[3] = 6;
assert(matrix[2, 3] == 6); // D & C index order
assert(matrix(3, 2) == 6); // Math & Fortran index order
}
// relaxed example
unittest
{
static Slice!(3, ubyte*) movingWindowByChannel
(Slice!(3, ubyte*) image, size_t nr, size_t nc, ubyte delegate(Slice!(2, ubyte*)) filter)
{
import std.algorithm.iteration : map;
import std.array : array;
auto wnds = image
.pack!1
.windows(nr, nc)
.unpack
.transposed!(0, 1, 4)
.pack!2;
return wnds
.byElement
.map!filter
.array
.sliced(wnds.shape);
}
static T median(Range, T)(Range r, T[] buf)
{
import std.algorithm.sorting : topN;
size_t n;
foreach (e; r)
buf[n++] = e;
auto m = n >> 1;
buf[0 .. n].topN(m);
return buf[m];
}
import std.conv : to;
import std.getopt : getopt, defaultGetoptPrinter;
import std.path : stripExtension;
auto args = ["std"];
uint nr, nc, def = 3;
auto helpInformation = args.getopt(
"nr", "number of rows in window, default value is " ~ def.to!string, &nr,
"nc", "number of columns in window default value equals to nr", &nc);
if (helpInformation.helpWanted)
{
defaultGetoptPrinter(
"Usage: median-filter [<options...>] [<file_names...>]\noptions:",
helpInformation.options);
return;
}
if (!nr) nr = def;
if (!nc) nc = nr;
auto buf = new ubyte[nr * nc];
foreach (name; args[1 .. $])
{
auto ret =
movingWindowByChannel
(new ubyte[300].sliced(10, 10, 3), nr, nc, window => median(window.byElement, buf));
}
}
@safe @nogc pure nothrow unittest
{
import std.algorithm.comparison : equal;
import std.range : iota;
immutable r = 1000.iota;
auto t0 = r.sliced(1000);
assert(t0.front == 0);
assert(t0.back == 999);
assert(t0[9] == 9);
auto t1 = t0[10 .. 20];
assert(t1.front == 10);
assert(t1.back == 19);
assert(t1[9] == 19);
t1.popFront();
assert(t1.front == 11);
t1.popFront();
assert(t1.front == 12);
t1.popBack();
assert(t1.back == 18);
t1.popBack();
assert(t1.back == 17);
assert(t1.equal(iota(12, 18)));
}
pure nothrow unittest
{
import std.algorithm.comparison : equal;
import std.array : array;
import std.range : iota;
auto r = 1000.iota.array;
auto t0 = r.sliced(1000);
assert(t0.length == 1000);
assert(t0.front == 0);
assert(t0.back == 999);
assert(t0[9] == 9);
auto t1 = t0[10 .. 20];
assert(t1.front == 10);
assert(t1.back == 19);
assert(t1[9] == 19);
t1.popFront();
assert(t1.front == 11);
t1.popFront();
assert(t1.front == 12);
t1.popBack();
assert(t1.back == 18);
t1.popBack();
assert(t1.back == 17);
assert(t1.equal(iota(12, 18)));
t1.front = 13;
assert(t1.front == 13);
t1.front++;
assert(t1.front == 14);
t1.front += 2;
assert(t1.front == 16);
t1.front = 12;
assert((t1.front = 12) == 12);
t1.back = 13;
assert(t1.back == 13);
t1.back++;
assert(t1.back == 14);
t1.back += 2;
assert(t1.back == 16);
t1.back = 12;
assert((t1.back = 12) == 12);
t1[3] = 13;
assert(t1[3] == 13);
t1[3]++;
assert(t1[3] == 14);
t1[3] += 2;
assert(t1[3] == 16);
t1[3] = 12;
assert((t1[3] = 12) == 12);
t1[3 .. 5] = 100;
assert(t1[2] != 100);
assert(t1[3] == 100);
assert(t1[4] == 100);
assert(t1[5] != 100);
t1[3 .. 5] += 100;
assert(t1[2] < 100);
assert(t1[3] == 200);
assert(t1[4] == 200);
assert(t1[5] < 100);
--t1[3 .. 5];
assert(t1[2] < 100);
assert(t1[3] == 199);
assert(t1[4] == 199);
assert(t1[5] < 100);
--t1[];
assert(t1[3] == 198);
assert(t1[4] == 198);
t1[] += 2;
assert(t1[3] == 200);
assert(t1[4] == 200);
t1[] *= t1[];
assert(t1[3] == 40000);
assert(t1[4] == 40000);
assert(&t1[$ - 1] is &(t1.back()));
}
@safe @nogc pure nothrow unittest
{
import std.range : iota;
auto r = (10_000L * 2 * 3 * 4).iota;
auto t0 = r.sliced(10, 20, 30, 40);
assert(t0.length == 10);
assert(t0.length!0 == 10);
assert(t0.length!1 == 20);
assert(t0.length!2 == 30);
assert(t0.length!3 == 40);
}
pure nothrow unittest
{
import std.experimental.ndslice.internal : Iota;
import std.meta : AliasSeq;
import std.range;
import std.typecons : Tuple;
foreach (R; AliasSeq!(
int*, int[], typeof(1.iota),
const(int)*, const(int)[],
immutable(int)*, immutable(int)[],
double*, double[], typeof(10.0.iota),
Tuple!(double, int[string])*, Tuple!(double, int[string])[]))
foreach (n; Iota!(1, 4))
{
alias S = Slice!(n, R);
static assert(isRandomAccessRange!S);
static assert(hasSlicing!S);
static assert(hasLength!S);
}
immutable int[] im = [1,2,3,4,5,6];
auto slice = im.sliced(2, 3);
}
pure nothrow unittest
{
auto tensor = new int[3 * 4 * 8].sliced(3, 4, 8);
assert(&(tensor.back.back.back()) is &tensor[2, 3, 7]);
assert(&(tensor.front.front.front()) is &tensor[0, 0, 0]);
}
pure nothrow unittest
{
import std.experimental.ndslice.selection : pack;
auto slice = new int[24].sliced(2, 3, 4);
auto r0 = slice.pack!1[1, 2];
slice.pack!1[1, 2][] = 4;
auto r1 = slice[1, 2];
assert(slice[1, 2, 3] == 4);
}
pure nothrow unittest
{
auto ar = new int[3 * 8 * 9];
auto tensor = ar.sliced(3, 8, 9);
tensor[0, 1, 2] = 4;
tensor[0, 1, 2]++;
assert(tensor[0, 1, 2] == 5);
tensor[0, 1, 2]--;
assert(tensor[0, 1, 2] == 4);
tensor[0, 1, 2] += 2;
assert(tensor[0, 1, 2] == 6);
auto matrix = tensor[0 .. $, 1, 0 .. $];
matrix[] = 10;
assert(tensor[0, 1, 2] == 10);
assert(matrix[0, 2] == tensor[0, 1, 2]);
assert(&matrix[0, 2] is &tensor[0, 1, 2]);
}