Humid air example fails due to new limits

[jowr]

The recently introduced limits for the HAPropsSi cause the examples to fail ungracefully. This should be fixed be fore we release 6.3.0.

[ibell]

I guess this also borks the verification code: http://www.coolprop.org/dev/fluid_properties/HumidAir.html#verification-script ?

[jowr]

Are you aware of the fact that 30% of the calls in the verification script yield "NaN" or a difference in psi_w larger than 1e-8?

[ibell]

:( no I didn't know that. Is that before the fixes to the boundaries?

…

On Fri, Apr 19, 2019, 12:28 AM Jorrit Wronski ***@***.***> wrote: Are you aware of the fact that 30% of the calls in the verification script yield "NaN" or a difference in psi_w larger than 1e-8? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1820 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAOGC25OG7FAWLQ7XDGDOCTPRFYBXANCNFSM4HFZZ6UQ> .

[jowr]

v6.2.1 yields 24% with errors or large deviations, after adding the limits we reached 31%.

[ibell]

How about 6.3 without the limit fix? I thought that was 100% success rate?

…

On Fri, Apr 19, 2019, 12:54 PM Jorrit Wronski ***@***.***> wrote: v6.2.1 yields 24% with errors or large deviations, after adding the limits we reached 31%. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1820 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAOGC24BTGEWSYEUCLRYZIDPRIPNPANCNFSM4HFZZ6UQ> .

[ibell]

I pushed a fix to the limits function (max and min were flipped for the default), but some issues remain. When I disable limits checking, I get zero failures again.

…

On Fri, Apr 19, 2019, 4:08 PM Ian Bell ***@***.***> wrote: How about 6.3 without the limit fix? I thought that was 100% success rate? On Fri, Apr 19, 2019, 12:54 PM Jorrit Wronski ***@***.***> wrote: > v6.2.1 yields 24% with errors or large deviations, after adding the > limits we reached 31%. > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#1820 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AAOGC24BTGEWSYEUCLRYZIDPRIPNPANCNFSM4HFZZ6UQ> > . >

[jowr]

Two things: - There is a branch for this issue with many fixes. Please do not edit the master. - I am not sure what you mean with success rate. As far as I can see you only check for exceptions and not for faulty results. That is where the problem is. As I said, I checked v6.2.1, which is without the limits and it also yields over 20% percent of wrong results. Neither the branch “issues/1820” nor v6.2.1 produce any exceptions.

[ibell]

Argh, sorry, I didn't realize there was a branch for this issue, should have looked into that. I'll work on the branch in the future. But on the question of the verification script, I'm rather confused. I added a check in the script to throw exceptions (in Python) when the value is not a valid number. That still throws no exceptions for me...

…

On Sat, Apr 20, 2019 at 2:24 AM Jorrit Wronski ***@***.***> wrote: Two things: - There is a branch for this issue with many fixes. Please do not edit the master. - I am not sure what you mean with success rate. As far as I can see you only check for exceptions and not for faulty results. That is where the problem is. As I said, I checked v6.2.1, which is without the limits and it also yields over 20% percent of wrong results. Neither the branch “issues/1820” nor v6.2.1 produce any exceptions. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1820 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAOGC23AZDZXIDQSFPQDGK3PRLHKVANCNFSM4HFZZ6UQ> .

[jowr]

You can check the CPP example in the “dev/tickets” folder. There is a counter for faulty results that are neither NAN nor exceptions.

[ibell]

Indeed I see what you mean. Boy these tests take forever ...

…

On Sat, Apr 20, 2019, 10:11 AM Jorrit Wronski ***@***.***> wrote: You can check the CPP example in the “dev/tickets” folder. There is a counter for faulty results that are neither NAN nor exceptions. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1820 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAOGC22Y5UNOEMCPFWPVFXDPRM6ELANCNFSM4HFZZ6UQ> .

[ibell]

All fixed. Tests are faster now too.