tests failing in MSYS2

[jdpipe]

I'm seeing various errors when running 'pytest' in solsticepy in (a fully updated version of) MSYS2. Below is some output. The key bit seems to be C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input.yaml:56068:32: invalid floating point number np.float64(7.115741913664257)'.`
This suggests we might be incorrectly producing the input yaml file, possibly due to recent changes in the numpy library...?

john@windo MINGW64 ~/solstice-scripts/tests
$ pytest -s
================================ test session starts =================================
platform win32 -- Python 3.12.9, pytest-8.3.5, pluggy-1.5.0
rootdir: C:/msys64/home/john/solstice-scripts
plugins: dependency-0.6.0
collected 11 items

test_cal_field.py DM np.float64(14.142135623730951)
dRmnp.float64(12.247089450151003)
Zone 0 Nrow 6 Nhel 35
Zone 1 Nrow 12 Nhel 70
Zone 2 Nrow 25 Nhel 140
Zone 3 Nrow 51 Nhel 280
Zone 4 Nrow 102 Nhel 560

Denest field 75950

Expanded field 22640
rm: cannot remove '*.csv': No such file or directory
.
test_cal_layout.py DM np.float64(14.142135623730951)
dRmnp.float64(12.247089450151003)
Zone 0 Nrow 6 Nhel 35
Zone 1 Nrow 12 Nhel 70
Zone 2 Nrow 25 Nhel 140
Zone 3 Nrow 51 Nhel 280
Zone 4 Nrow 102 Nhel 560

Denest field 75950
0.0 7.2 0.0
90.0 4.4085828737605584e-16 7.2
180.0 -7.2 8.817165747521117e-16

Expanded field 22640
22640
.DM np.float64(14.142135623730951)
dRmnp.float64(12.247089450151003)
Zone 0 Nrow 6 Nhel 35
Zone 1 Nrow 12 Nhel 70
Zone 2 Nrow 25 Nhel 140
Zone 3 Nrow 51 Nhel 280
Zone 4 Nrow 102 Nhel 560

Denest field 75950

Expanded field 22640
.
test_cal_sun.py .rm: cannot remove '*.csv': No such file or directory
.
test_design_crs.py 0.5
120.0
Case directory is 'C:/msys64/home/john/solstice-scripts/tests/test-crs-design'
DM np.float64(14.142135623730951)
dRmnp.float64(12.247089450151003)
Zone 0 Nrow 4 Nhel 22
Zone 1 Nrow 8 Nhel 44
Zone 2 Nrow 16 Nhel 88
Zone 3 Nrow 32 Nhel 176
Zone 4 Nrow 64 Nhel 352

Denest field 30008

Expanded field 2800
attenuation 1e-06
Generating YAML file...

Start field design

Sun position: 1
azimuth: 270.00 , elevation: 29.11
Running 'solstice' with args: -D270.0,29.11 -v -t 1 -n 1000000 -R C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input-rcv.yaml -fo ./test-crs-design/sunpos_1/simul C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input.yaml
C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input.yaml:56068:32: invalid floating point number `np.float64(7.115741913664257)'.
        by C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input.yaml:56068:18
        by C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input.yaml:56068:5
        by C:/msys64/home/john/solstice-scripts/tests/test-crs-design/input.yaml:56066:3
F
test_design_dish.py sCase directory is 'C:/msys64/home/john/solstice-scripts/tests/test-dish-parabolia'

[yewang726]

The issue is due to the changes in scalar string representation in Numpy, specifically (Numpy 1.20+). Starting around NumPy 1.20.0 (released January 2021), NumPy changed how its scalar types (like np.float64) are displayed in containers (lists, tuples, etc).

Explicitly converting NumPy scalars to native Python types in the YMAL generation function is the solution.

[jdpipe]

I found that the issue more specifically the use of '%s' string substitution for lists of values, eg

'...%s...' % ([val1,val2,np.float(npval3)])

What previously gave you stringified floats now gives you stringified code instead ie instead of [1,2,3.14159] you now get [1,2,np.float(pi)]. As you say, there was a numpy change that caused this.

I think it's fairly readily fixed. We should be using repeated %e or eg %6.8f string substitutions for vectors rather than the current approach. We should never be using %s substitution when we actually want numerical values (or lists) inserted.

[yewang726]

The issue is solved. See the details in the pull request #4: #4