Input string for the Fit algorithm get rejected when presented with a unix-style path

[Martyn Gigg]

The attached script and nexus files should allow replication of the problem.

The debugger shows that an exception is thrown in API::Expression::tokenize. I think it is to do with "/" being treated as an operator(divide) so the string parsing falls over.

Note: This is a unix-style path issue so only occurs on Linux/Mac

[Martyn Gigg]

(In [8181]) Fix a problem with unix-style paths in building a string representation of the resolution function. Fixes #2090

[Martyn Gigg]

(In [8182]) Clear some warnings with the IDA interface. Fixes #2090

[Michael Reuter]

I tried running the attached script from within MantidPlot (adjusting the paths for where I downloaded the data) and it doesn't work. Here's the error I see:

Fit-[Error] Error in execution of algorithm Fit:
Fit-[Error] Error in input string to FunctionFactory
name=LinearBackground,A0=0,A1=0,ties=(A0=0.00,A1=0.0);(composite=Convolution;name=Resolution,FileName=/home/2zr/Downloads/irs26173_graphite002_res.nxs;name=Lorentzian,Height=29.1171,PeakCentre=-0.00181586,HWHM=0.000406624)

It looks like the same problem, but could be something new.

[Martyn Gigg]

So

[Martyn Gigg]

So the problem here was with the script generated by the algorithm history and not with the fitting within the IDA interface.

The issue is of having strings inside strings and quoting them correctly. If they are backslashed appropriately they will work. I've update the attachment to a working copy and raised a separate ticket about the more general issue of script generation in these cases. See #2101

[Michael Whitty]

with correctly escaped input script, this works in [8190] (latest version with rpm build available). also tested on the IDA interface.

[Stuart Campbell]

This ticket has been transferred to github issue ​2937