The attached zip has two ibn waves. One with standard NaNs and one with NaNs from CurveFit.
These give different results when serialized:
•print zeromq_test_serializeWave(normalNaN)
{
"data": {
"raw": [
"nan",
"nan"
]
},
...
•print zeromq_test_serializeWave(strangeNaN)
{
"data": {
"raw": [
"-nan(ind)",
"-nan(ind)"
]
},
...
The "strange" NaN can be created with
•make/D xData = {1, 1}
•make/D yData = {10, 10}
•CurveFit line yData/X=xData
y= W_coef[0]+W_coef[1]*x
W_coef={NaN,NaN}
V_chisq= NaN;V_npnts= 2;V_numNaNs= 0;V_numINFs= 0;V_startRow= 0;
V_endRow= 1;V_q= 1;V_Rab= NaN;V_Pr= NaN;V_r2= NaN;
W_sigma={NaN,NaN}
Coefficient values ± one standard deviation
a = NaN ± NaN
b = NaN ± NaN
•print zeromq_test_serializeWave(w_coef)
I think we should always standardize the output to nan. Check also the same for inf.
someNaNs.zip
The attached zip has two ibn waves. One with standard NaNs and one with NaNs from CurveFit.
These give different results when serialized:
The "strange" NaN can be created with
I think we should always standardize the output to
nan. Check also the same forinf.someNaNs.zip