[Kst] Re: reading ASCII files / KST command line options

[rrebel at rfecllc.com]

Dear Nicolas,

thank you for your fast response. Meanwhile I got the command line
version to work. The problem is more or less the available documentation.
 
I checked out the repository to have a look at the source and voila,
there I found under design specs a document describing the command line
options for the 2.02 Version as well as all the documentation required
for the reader plug in. By searching the documents I found that the
order the options are given, matters. This is a little unusual so. First
the datafile and then the options in the right order. (-n in front tof
the plot definitions).
But on top of this you have to set the default settings for the reader
in a way that it can read the file without error. This means you have to
set the  delimiter, the data start line, the comment symbol's and the
line where the headers are as the default settings. Otherwise the reader
fails and no data are available. Unfortunately there are no command line
options to specify all that.
In my opinion the ASCII reader is to intollerant on the format.

The ".kst" file seems to be a real bug.

Anyway for more complex automation it would be nice that onealso could
specify a script on the command line to run. This would be the most
powerful and flexible way for automation since the program generating
the data could also generate the script for post processing and
displaying the data.

I also took a look on the reader plug in. I'm not an experienced C++
programmer but I think I can manage it. It is just a matter of priority
and time. Since the ASCII works now I have a solution for my problem and
it just sunk a little in importance.

Best Regards

Reimund Rebel




On 1/30/2011 14:19, Nicolas Brisset wrote:
> Dear Reimund,
>
>> I want to use KST as a general Plot and View Tool in a test lab
>> environment. For several reasons we decided to store the measured
>> data in simple CSV files. The data sets contain a header (embedded in
>> Comment lines) with some information about the measurement itself. Then a
>> line with the column headers followed by the dataset.  I attached a
>> sample.
> This should work without any problem. In the coming version (2.0.3) you will even be able to display the metadata contained in the header in your plots.
>
>> kst  -x 1 -y 3 -y 4 -n 500 -m 1  -g  test.csv     or
>> kst  -x Time  -y Temp  -y V1 -n 500 -F test.csv   setup1.kst    
>>
>> So far I had no luck with both of them. On a linux box,  I have the
>> feeling it does understand the command line but it seems to have
>> trouble reading the data in ending in blank graphs. On windows it behaves
>> even worse. It always shows just a blank screen and it looks like it does
>> not even read the setup.kst file
>>
>> Attached a sample file and the setup.kst
> I have just a done a quick test, and I can read the data with the data wizard. 
> However, I confirm that loading your .kst does not work.
> I don't know about the -F flag: Barth, is it still available with 2.0.x and if yes, has it been tested thoroughly?
> I think you've uncovered a bug, we'll try to fix it ASAP. We'd like to release 2.0.3 in the coming days, it would be nice to have solved it until then. By the way, I did the previous tests with today's svn, not 2.0.2...
>
>> By the way is it possible to read data also from stdin or a pipe,
>> this would be an elegant way to plot directly from the test program.
> I think it is possible, but I am not a stdin user so I'll let someone else give you a more detailed answer.
>
>> Long term I also would be interested to write a plugin for a
>> different (binary) file format for exchange with a free circuit simulation tool
>> (Switcher CAD ).
> Peter already gave you some hints. Basically, you need to be able to:
> - detect the file type and announce that you can read it
> - list the variables it contains 
> - provide the number of samples for each variable
> - provide the values for a given range of samples
> - if the file grows and you want to plot "live" curves, you'll need to implement the corresponding update() mechanism
> The best way is probably trying to do it, and reporting here the issues you run into and the questions you have. Be aware that the ASCII datasource is a pretty capable and heavily optimized piece of code. I don't know how good a coder you are, but if you're just a hobbyist as I am, you might get scared. NetCDF is much easier to understand, even though it has quite some stuff for metadata and primitive types (vectors, matrices, strings and scalars are so-called "primitives") which you may not need to support.
>
> And if I can recommend something: use QtCreator to develop that, it'll help you a lot!
>  
> Best regards,
>
> Nicolas