This is a list of some tasks that were not completed or not even started.

They differ from the list of ideas in the sense that these should have a higher importance and/or impact in the user experience, according to the existing features.

Of course, even more important are the known bugs and limitations.

This list follows no specific order or categories.

TODO

1. Add an option in cost.py for absolute or relative counting of days/months.
2. Check/fix if convert_gslib is creating the file even when it is not going to write it.
3. Try to automate the numbers formatting when writing files.
4. [Refactor] Remove references to stations merging.
5. Confirm if the DSS binary is floor'ing variograms' range. Act accordingly.
6. Option to append year (maybe this?.
7. Integrate keys in the overall process.
8. Deal with stations outside of the grid. Check grid limits and put them aside?
9. [Refactor] Review the need for flag (optional or mandatory?)
10. Confirm the desired behaviour of this call to dsspar.update.
11. Check the exclusions when using the option to skip outliers in the scores calculation:
    • station: exclude the year
    • network: exclude the station
12. Validate rows (scores GUI), maybe with QValidator.
13. Test/fix homogenisation with one single station.
14. Maybe add an option in cost.py to append a column with the year when resolution != 'y' or when sum_year == True.
15. Check/fix station_order when header == False.
16. Fix the hack in flush_varnames related with the 'flag'.
17. Make summary of the merge_output compatible to when not using batch.
18. Check/fix fill_station (here and here) for non annual data. Also, why do we have two functions? There was some refactoring lost in the way...
19. [Refactor] Merge loadcheck with PointSet.load, or to the parent class.
20. Omit results per decade in the results.xls of the batch decade.
21. Add option to do not fill no_data, or to fill it with other value (e.g., average or user defined). Make it independent, by first inserting no_data values where needed.
22. Handle variography definition when using batch networks without batch decade.
23. save_settings after loading does not update the original file. Maybe ask when closing.
24. Check filling stations in monthly data.
25. Add buttons to change tab.
26. Create a gsimcli_results.xls file with a single hard data.
27. There are other TODO's spread over the code
28. Most likely there is a lot of code cleaning to do (including removing commented out blocks that are not useful anymore).

This is a list of some new features and ideas that I had, which were never implemented or tried out.

They differ from the to do list in the sense that those should have a higher importance and/or impact in the user experience, according to the existing features.

Of course, even more important are the known bugs and limitations.

They are grouped into two categories:

• Method: related with the methodology itself.
• Application: strictly related with the software.

Method

1. Test using octants in DSS (Does simulation work? Is it only for two part search?).
2. Add an test the option to crop the first n years. It is already implemented here, but not in the front-end. Venema (2012) excludes the first and last 5 years (see page 95).
3. Implement the score power of detection (Venema, 2012).
4. Test smaller search radius.
5. Test candidate as soft data.
6. Add and test mask/buffer. It should highly reduce the simulation time.
7. Implement the score CMRSE Anomalies.
8. Try DSS version with local means.
9. Test PCA/clusters.
10. Calculate and flag abrupt changes between corrected and observed values, e.g., column with (obs-corr)/corr.

Application

1. More outputs:
   • histograms
   • maps
   • other graphics
   • visualize variograms
2. Class to encapsulate/manage all the files of the gsimcli procedure (input, output, intermediary, parameters).
3. Convert between file types:
   • GSLIB to COST-HOME
   • COST-HOME to CSV/gsimcli
   • GSLIB to CSV
   • CSV to GSLIB
   • GSLIB add/remove header
4. Ask the user the number of divisions over ZZ.
5. Add functions to split hard data.
6. Handle different versions of the DSS (running and reading/loading parameters).
7. Add a cute QSplashScreen.
8. Save results by station.
9. Show simstats progress.
10. Handle simstats parameters.
11. Try to use multiprocessing to detect irregularities.
12. Try optimizations (Cython, Numba, PyPy, Pyston).
13. Maybe replace parameters handling with ConfigParser
14. Load/save data/parameters in binary format.
15. Improve clearing the GUI settings, default/reset.
16. Estimate remaining time.
17. Provide quick launch/settings.
18. Handle user errors/missing parameters.
19. Load/save grid and variog files.
20. Show variog file in batch decades.
21. mp_exec with symlink instead of copy (not sure if it would work).
22. Convert to COST-HOME when finalizing batch decade.
23. To/from gsimcliparam.
24. Allow gsimcli_results without batch_decade.
25. Develop a CLI version, including docopt.
26. Bundle an .exe (cx_freeze, PyInstaller, Cython, Nuitka).

Here is a list of known bugs and limitations. I grouped them into two categories:

• Major relevance: the user has to be careful about these.
• Minor relevance: small issues that should not be noticeable to the user, and may be in parts of the code that are not being used in the last version.

There is no particular order within the categories.

Major relevance

1. Does not support file paths with :/ (which can happen in POSIX systems).

2. In some unknown occasions, using the menu option to "restore/load settings" may add a duplicate network (double check the list and remove as needed).

3. Make sure all the candidate stations are within the grid limits, otherwise it will throw an error while trying to extract the corresponding vertical line.

4. Related with the previous point, be careful to define the grid dimensions in order to accomodate the radius (tolerance) around the stations. Example:

   • the grid is 10 x 10 nodes
   • one station in located at node (2, 2)
   • if using tolerance 3, it will surpass the grid dimensions in the west and north sides
   • you need to increase the grid to at least 11 x 11 (in the aforementioned directions)

5. When calculating the scores for monthly data in the cost-home format, it will only use the first selected checkbox.

Minor relevance

1. cost2gslib only adds the column with the 'year' for annual and monthly data.
2. drill has some offset in the fetched location (not in use).
3. GsimcliParam.load is being used like it was a static method (it should not be a problem because most likely the first condition will short circuit).
4. Using the option to save intermediary files, coupled with using a tolerance, may throw an error while saving.
5. benchmark is not updating the header when using 'monthly'.