[FEATURE REQUEST] numpy.delete

[water5]

https://numpy.org/doc/stable/reference/generated/numpy.delete.html

I think this is a basic function.

[v923z]

But can't this functionality be reproduced with slicing, or compress? https://numpy.org/doc/stable/reference/generated/numpy.compress.html compress is implemented in the yet unmerged PR #456.

[water5]

In some use case, numpy.compress is better than numpy.delete, but when just need to delete several rows/columns of a big array, likes below:

b = np.arange(10000).reshape(100, 100)

array([[ 0, 1, 2, ..., 97, 98, 99],
[ 100, 101, 102, ..., 197, 198, 199],
[ 200, 201, 202, ..., 297, 298, 299],
...,
[9700, 9701, 9702, ..., 9797, 9798, 9799],
[9800, 9801, 9802, ..., 9897, 9898, 9899],
[9900, 9901, 9902, ..., 9997, 9998, 9999]])

np.delete(b, 98, axis = 0)

array([[ 0, 1, 2, ..., 97, 98, 99],
[ 100, 101, 102, ..., 197, 198, 199],
[ 200, 201, 202, ..., 297, 298, 299],
...,
[9600, 9601, 9602, ..., 9697, 9698, 9699],
[9700, 9701, 9702, ..., 9797, 9798, 9799],
[9900, 9901, 9902, ..., 9997, 9998, 9999]])

np.delete(b, 98, axis = 1)

array([[ 0, 1, 2, ..., 96, 97, 99],
[ 100, 101, 102, ..., 196, 197, 199],
[ 200, 201, 202, ..., 296, 297, 299],
...,
[9700, 9701, 9702, ..., 9796, 9797, 9799],
[9800, 9801, 9802, ..., 9896, 9897, 9899],
[9900, 9901, 9902, ..., 9996, 9997, 9999]])

If use numpy.compress, need to construct a 1-D array of bools first use for condition, when origin array huge, likes above case, it's not a easy job?

[v923z]

In some use case, numpy.compress is better than numpy.delete, but when just need to delete several rows/columns of a big array, likes below:

If use numpy.compress, need to construct a 1-D array of bools first use for condition, when origin array huge, likes above case, it's not a easy job?

This is a good point, I completely overlooked that. We can definitely add delete. There are a couple of other functions in the pipeline, but I should be able to pull this off in a week.

[water5]

Thanks.

[v923z]

@water5 Could you, please, check out #473? I think the code should work (the tests passed). I haven't implemented indexing via slices, so you can do something like this:

np.delete(a, 1, axis=0)
np.delete(a, [1, 2, 3, 4], axis=0)
np.delete(a, (1, 2, 3, 4), axis=0)
np.delete(a, range(5), axis=0)

(Actually, any iterable should do in place of the indices, so you can simulate slices via the range function.)

Also note that the indices are copied once, because I have to sort them in increasing order. That means that 4*len(indices) bytes are required to store the values temporarily. I don't know if that is acceptable, but I don't actually see another option, since there is no guarantee that the indices are sorted.

[water5]

When just delete 1 row/column,
must use np.delete(a, [row/column], axis = 0) or np.delete(a, (row/column, ), axis = 0) or np.delete(a, [5, ], axis = 0),

np.delete(a, row/column, axis = 0) or np.delete(a, (row/column), axis = 0) not work.

a = np.arange(100).reshape((10, 10))
np.delete(a, 5, axis = 0)

not work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[50, 51, 52, 53, 54, 55, 56, 57, 58, 59],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

np.delete(a, (5), axis = 0)

not work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[50, 51, 52, 53, 54, 55, 56, 57, 58, 59],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

np.delete(a, (5, ), axis = 0)

work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

np.delete(a, [5], axis = 0)

work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

np.delete(a, [5, ], axis = 0)

work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
[30, 31, 32, 33, 34, 35, 36, 37, 38, 39],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[70, 71, 72, 73, 74, 75, 76, 77, 78, 79],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

When delete multi rows/columns,
use np.delete(a, (3, 5, 7), axis = 0) or np.delete(a, [3, 5, 7], axis = 0) both work,

np.delete(a, (3, 5, 7), axis = 0)

work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

np.delete(a, [3, 5, 7], axis = 0)

work,

array([[0, 1, 2, 3, 4, 5, 6, 7, 8, 9],
[10, 11, 12, 13, 14, 15, 16, 17, 18, 19],
[20, 21, 22, 23, 24, 25, 26, 27, 28, 29],
[40, 41, 42, 43, 44, 45, 46, 47, 48, 49],
[60, 61, 62, 63, 64, 65, 66, 67, 68, 69],
[80, 81, 82, 83, 84, 85, 86, 87, 88, 89],
[90, 91, 92, 93, 94, 95, 96, 97, 98, 99]], dtype=int16)

---

Is it match design purpose?

[v923z]

When just delete 1 row/column, must use np.delete(a, [row/column], axis = 0) or np.delete(a, (row/column, ), axis = 0) or np.delete(a, [5, ], axis = 0),

np.delete(a, row/column, axis = 0) or np.delete(a, (row/column), axis = 0) not work.

Thanks for looking into this. This is indeed an error, I will fix it as soon as possible.

[v923z]

@water5 Could you, please, check out the latest commit in #473?

[water5]

Thanks, all test result correct currently.

By the way, Question about git use,

git clone https://github.com/v923z/micropython-ulab.git
cd micropython-ulab
git checkout io

Branch 'io' set up to track remote branch 'io' from 'origin'.
Switched to a new branch 'io'

git pull

Already up to date.

Why can't refresh files and appear 'Already up to date.'? Which step I miss?
I use download / extract zip file at this time.

[v923z]

Thanks, all test result correct currently.

Many thanks for the feedback!

By the way, Question about git use,

git clone https://github.com/v923z/micropython-ulab.git
cd micropython-ulab
git checkout io

Branch 'io' set up to track remote branch 'io' from 'origin'.
Switched to a new branch 'io'

git pull

Already up to date.

Why can't refresh files and appear 'Already up to date.'? Which step I miss? I use download / extract zip file at this time.

You should usually fetch the content before pulling

git fetch
git pull

[water5]

Build error with STM32 and RP2, but build pass with ESP32. (master branch)
make BOARD=ADAFRUIT_F405_EXPRESS USER_C_MODULES=../../../micropython-ulab all
make BOARD=PICO USER_C_MODULES=../../../micropython-ulab/code/micropython.cmake

CC ../../../micropython-ulab/code/numpy/stats.c
CC ../../../micropython-ulab/code/numpy/transform.c
../../../micropython-ulab/code/numpy/transform.c: In function 'transform_delete':
../../../micropython-ulab/code/numpy/transform.c:209:9: error: unknown type name 'ssize_t'; did you mean '_ssize_t'?
209 | ssize_t value = (ssize_t)mp_obj_get_int(indices);
| ^~~~~~~
| _ssize_t
../../../micropython-ulab/code/numpy/transform.c:209:26: error: 'ssize_t' undeclared (first use in this function); did you mean '_ssize_t'?
209 | ssize_t value = (ssize_t)mp_obj_get_int(indices);
| ^~~~~~~
| _ssize_t
../../../micropython-ulab/code/numpy/transform.c:209:26: note: each undeclared identifier is reported only once for each function it appears in
../../../micropython-ulab/code/numpy/transform.c:209:34: error: expected ',' or ';' before 'mp_obj_get_int'
209 | ssize_t value = (ssize_t)mp_obj_get_int(indices);
| ^~~~~~~~~~~~~~
../../../micropython-ulab/code/numpy/transform.c:213:45: error: expected ')' before 'axis_len'
213 | if((value < 0) || (value > (ssize_t)axis_len)) {
| ~ ^~~~~~~~
| )
../../../micropython-ulab/code/numpy/transform.c:222:20: error: expected ';' before 'value'
222 | ssize_t value = (ssize_t)mp_obj_get_int(item);
| ^~~~~~
| ;
../../../micropython-ulab/code/numpy/transform.c:223:16: error: 'value' undeclared (first use in this function)
223 | if(value < 0) {
| ^~~~~
../../../micropython-ulab/code/numpy/transform.c:226:49: error: expected ')' before 'axis_len'
226 | if((value < 0) || (value > (ssize_t)axis_len)) {
| ~ ^~~~~~~~
| )
make: *** [../../py/mkrules.mk:77: build-ADAFRUIT_F405_EXPRESS/code/numpy/transform.o] Error 1

[v923z]

@water5 I think a header file was missing. Could you, please, check out https://github.com/v923z/micropython-ulab/tree/delete-fix to see if that fixes the issue?