source: sasview/guitools/requirements.txt @ 1db4a53

- BUGS

  1- [DONE] After fitting, the fit panel is not updated with the output values of the fit. 
  
  2- [DONE] Do this:
      a) load the file line_in_x2_y.txt
      b) fit a straight line, close the Fit Dialog
      c) change the scale using the "Properties" menu to x vs y (linear vs linear)
      You will see an exception that should be fixed.

  3-  [DONE] Load the file line_in_x2_y.txt in miniView. Select x^2 and y for the scales. Fit it. 
      a) The fit range is wrong, it's not update for the X^2 view.
      b) the fit result is wrong
      c) Even if you update the range, the fit and the plots are wrong.
      
  4- [DONE]At the python command prompt, try:
       from sans.guitools.PlotPanel import PlotPanel
       
       You will get an exception: ImportError: No module named fittings
       
     Guitools should not depend on your miniView code.
     
  5- [QUESTION] Each time you open and close the Fit Dialog, a new label "Theory1D" is created in the
     legend. There should always be only one, for the _last_ fit, and the name should be "Fit".
     [alina] is it fine if I delete the fit line from the graph
      and remove  it when I close the Fit Dialog window?
     [Mathieu] That might work. I kind of like it. Let's do it like that and see what the users
      need. 
     
  6-[QUESTION ] Do this:
                a) Load the file 'line_in_logy_x.txt'
                b) Fit a straight line
                c) The fit shown does not start from the first point and it's not a straight line. 
        [ALINA] then for transformation to log Do I have to transform x to log and set x linear?
        [Mathieu] No. That would confuse the user, who expects a log scale. The answer here is to fit 
         y = A * log(x) + B, which is a straight line in the y vs log(x) representation.
         
--> [ALINA] woring better...I am concerned how to fit a straight line when x is changing too?
         
      
- NOTES that might help for bugs
  1- Make sure you are fitting the View data, not the underlying data
  2- Since you are using the log scale from matplotlib, the fit for when one of 
                the axis is in log will NOT be a straight line
  3- Make sure that you are filling the View of the fit result correctly.
  [ALINA] not using the log of matplotlib anymore but the axis are never in the log scale by the value are in log10 x
  [Mathieu] See my comment to point 6. The right answer is to go back to using the log scale of matplotlib.

- NEEDED IMPROVEMENTS
  1-[DONE] After loading a new file or changing the scale using the "Properties" menu, remove any previously displayed fit from the graph.
  
  2- There are additional y scales we need: y, y^2, log(y), 1/y, 1/sqrt(y), log(y * x), log(y * x^2)
  [QUESTION] when plottong y=log(y*x) is the user allowed to do transformation on x at the same time 
  for example x---> log X with  y=log(x*y)
  [PROBLEM]WITH  log(y * x), log(y * x^2)
        
  3-[DONE] You should change the label according to what the user chose, and not hard-code them.
     For example, in PlotPanel._onEVT_FUNC_PROPERTY, replace
          
            self.graph.xaxis('\\rm{q^{2}} ', 'A^{-2}')
         
         with
            name, units = self.some_plottable.get_xaxis()
            self.graph.xaxis("(%s)^2" % name,  "(%s)^2" % units)
    
     That way it still works even if the user is not plotting Intensity versus Q.
     
  4- [DONE] In the fit dialog, put the "fit" and "Close" button in the bottom right corner, like in a standard file dialog.
  
  5- The layout of the Fit Dialog should be improved.
  
  6- The code should be well documented and clean-up (no print statements).
  
  
  
  AFTER TALKING WITH STEEVE KLEIN
  
 This is a list the changes Steeve wants :
 
  1- [DONE] Add curves : ln(y) in linear scale, ln(x^(4)*I)
  
  2- [DONE]Specify in the fitDialog if natural log or log base 10
  
  3- [FIXED]When fitting sphere.txt , logx --logy ,fitting is not correct
  
  4- [FIXED]the line to fit always is displayed from x min to max all the time even if xmin is changed
     or xmax is changed
     
  5- [FIXED]The fitting errors are too small (order of 10^(-7))for that case too,certainly wrong
  
  6- When to see errors bars appear when loading sphere.txt file. 
  
  7- Add zoom in option when xmax and xmin(user) are entered.
  
  8-[FIXED] Always want to know the initial value of xmin and xmax when opening the fitDialog.These values
     should always be independant of the transformation
     
  9-[FIXED] the user should be able to add point to fit without knowing the exact range or guessing
                a range to fit.

        TO Dr. MATHIEU 
  1- I probably have to change the interface of the fitdialog, I would like to discuss 
     it before proceeding
        ALINA
 1- Would like to change the y axis label display
 
 
Cleaning up for deployment:

 - When zooming, check that the new lower bound is positive if on a log scale [DONE] 
 - Fixed bug with the zooming dependency on scale [DONE]
 - Improved look of Fit dialog [DONE]
 
 - Improve labels
 
 - Problem scales:
     - x vs 1/y