Extend broadcasting rules of numpy to abstract tabularshape of data from cellular shape
Project description
Introduction
Some engineering problems strain array representation due to tabulation and degeneracy. tablarray disambiguates data structure of tables-of-cells and extends powerful methods into these domains.
Linked below are introductions to the concepts:
the Big Idea of TablArray is to disambiguate tables of cells, which traditionally would have been stacks of arrays, or arrays of arrays. In this doc, we see that new math operators are provided with upgraded broadcasting rules, and in examples we see that we can now write formulas with a blind eye toward tabular convention as well as parameter degeneracy.
the Big Idea of TablaSet (work in progress) is to extend TablArray into a territory that overlaps roughly with a database, and yet we retain the speed of slicing. In fact, when we marry database and slicing concepts, a new concept emerges, called projection. As a side benefit, TablaSet also has some elegant utilities (or will soon).
the Big Idea of TablaSolve (work in progress) is to build object-oriented datasets by adding computational flow on top of a TablaSet. In practice, this separates a modeling task into defining operators that solve individual dependencies, defining a seed data-set, and then handing control to the TablaSolve to resolve all of the system’s dependent parameters.
(tablarray was originally developed to manage large numbers of optical modes in laser simulation.)
Brief Illustration
import numpy as np
import tablarray as ta
x = ta.TablArray(np.linspace(-2, 2, 4), 0)
y = ta.TablArray(np.linspace(-1.5, 1.5, 4).reshape(4,1), 0)
E = ta.zeros((4, 4, 2), cdim=1)
E.cell[0] = 1.5 / (x**2 + y**2)
E.cell[1] = -.01 + x * 0
print(E)[[|[ 0.24 -0.01 ]| |[ 0.557 -0.01 ]| |[ 0.557 -0.01 ]| |[ 0.24 -0.01 ]|] [|[ 0.353 -0.01 ]| |[ 2.16 -0.01 ]| |[ 2.16 -0.01 ]| |[ 0.353 -0.01 ]|] [|[ 0.353 -0.01 ]| |[ 2.16 -0.01 ]| |[ 2.16 -0.01 ]| |[ 0.353 -0.01 ]|] [|[ 0.24 -0.01 ]| |[ 0.557 -0.01 ]| |[ 0.557 -0.01 ]| |[ 0.24 -0.01 ]|]]t(4, 4)|c(2,)
Those ‘|’ separate tabular vs cellular structure. Similarly ‘t(4, 4)|c(2,)’ is a reminder that E is set of arrays ‘c(2,)’ arranged in a 4x4 table ‘t(4, 4)’.
En = ta.abs(ta.linalg.norm(E)**2)
Efield = ta.TablaSet(x=x, y=y, E=E, En=En)
Efield['r'] = ta.sqrt(Efield['x']**2 + Efield['y']**2)
ta.set_printoptions(threshold=10, precision=3)
print(Efield.table) | x | y | E | En | r |
--------+--------+--------+----------+-------+-------+
[0, 0] | -2.000 | -1.500 | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[0, 1] | -0.667 | | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[0, 2] | 0.667 | | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[0, 3] | 2.000 | | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[1, 0] | | -0.500 | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[1, 1] | | | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[1, 2] | | | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[1, 3] | | | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[2, 0] | | 0.500 | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[2, 1] | | | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[2, 2] | | | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[2, 3] | | | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[3, 0] | | 1.500 | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[3, 1] | | | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[3, 2] | | | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[3, 3] | | | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
print(Efield.bcast) | x | y | E | En | r |
--------+--------+--------+----------+-------+-------+
[0, 0] | -2.000 | -1.500 | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[0, 1] | -0.667 | -1.500 | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[0, 2] | 0.667 | -1.500 | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[0, 3] | 2.000 | -1.500 | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[1, 0] | -2.000 | -0.500 | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[1, 1] | -0.667 | -0.500 | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[1, 2] | 0.667 | -0.500 | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[1, 3] | 2.000 | -0.500 | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[2, 0] | -2.000 | 0.500 | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[2, 1] | -0.667 | 0.500 | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[2, 2] | 0.667 | 0.500 | [ 2.16 | 4.666 | 0.833 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[2, 3] | 2.000 | 0.500 | [ 0.353 | 0.125 | 2.062 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[3, 0] | -2.000 | 1.500 | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
[3, 1] | -0.667 | 1.500 | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[3, 2] | 0.667 | 1.500 | [ 0.557 | 0.310 | 1.641 |
| | | -0.01 ] | | |
--------+--------+--------+----------+-------+-------+
[3, 3] | 2.000 | 1.500 | [ 0.24 | 0.058 | 2.500 |
| | | -0.01] | | |
--------+--------+--------+----------+-------+-------+
print(Efield.cell) | x | y | E | En | r |
-----+------------+------------+-------------+-------------+-------------+
[0] | [|-2.| | [[|-1.5|] | [[|0.24| | [[|0.058| | [[|2.5| |
| |-0.667| | [|-0.5|] | |0.557| | |0.31| | |1.641| |
| | 0.667| | [| 0.5|] | |0.557| | |0.31| | |1.641| |
| | 2. |] | [| 1.5|]] | |0.24 |] | |0.058|] | |2.5 |] |
| | | [|0.353| | [|0.125| | [|2.062| |
| | | |2.16| | |4.666| | |0.833| |
| | | |2.16| | |4.666| | |0.833| |
| | | |0.353|] | |0.125|] | |2.062|] |
| | | [|0.353| | [|0.125| | [|2.062| |
| | | |2.16| | |4.666| | |0.833| |
| | | |2.16| | |4.666| | |0.833| |
| | | |0.353|] | |0.125|] | |2.062|] |
| | | [|0.24| | [|0.058| | [|2.5| |
| | | |0.557| | |0.31| | |1.641| |
| | | |0.557| | |0.31| | |1.641| |
| | | |0.24 |]] | |0.058|]] | |2.5 |]] |
-----+------------+------------+-------------+-------------+-------------+
[1] | | | [[|-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01|] | | |
| | | [|-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01|] | | |
| | | [|-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01|] | | |
| | | [|-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01| | | |
| | | |-0.01|]] | | |
-----+------------+------------+-------------+-------------+-------------+
Lessons from above:
TablArray and TablaSet have bcast, table, and cell views.
Broadcasting rules of numpy are extended to recognize tabular and cellular shapes.
This frees physics libraries to write formulas while blind to tabular super-structure of the application. In other words, the goal is to abstract formulas from tabular shape.
TablaSet adds to TablArray by enforcing broadcast-ability across datasets. Once a TablaSet is built, you know it is ready for formulas.
Installation
pip install tablarray
Status
Alpha - tablarray might be stable enough for prototype applications, though it may continue to be unstable for a while longer.
I.e.:
Critical features are implemented and not expected to change significantly.
A few features need further adaptation for certain cases.
A little testing is done, most not so much.
Some features are still missing.
I remain concerned about undesirable, or worse - undefined, behavior at edge cases.
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file tablarray-0.5.0.tar.gz.
File metadata
- Download URL: tablarray-0.5.0.tar.gz
- Upload date:
- Size: 46.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.11.2
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
33eeed0b39aed7b587ea85383c33d18f34e339f63665da7009cfdf13d277db83
|
|
| MD5 |
4eb32686e53814d817c78fcc65ef1ec6
|
|
| BLAKE2b-256 |
59b7da97a32429a9b5b5dc874c4497956397aedc1f1ebedd30c039bb2ae200ac
|
File details
Details for the file tablarray-0.5.0-py3-none-any.whl.
File metadata
- Download URL: tablarray-0.5.0-py3-none-any.whl
- Upload date:
- Size: 53.0 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/4.0.2 CPython/3.11.2
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
d625221a87df39b83bf876fff54e3e560a697a50041d26793288bd8d939273a9
|
|
| MD5 |
0fb95d6c0eed18c60426700ecbeeea6b
|
|
| BLAKE2b-256 |
9d7299e5225b822bc4f558afd0702f12256605e6da277dd15b9096139e568eb8
|
