Fix "//" in URL concatenation

[chris-hld]

When the URL of the specified database ends with an "/", the concatenation results in "//".

[fietew]

Does this result in failed downloads? If yes, which?

[hagenw]

I think @chris-hld reported it for example for https://github.com/TWOEARS/examples/tree/master/qoe_coloration

[fietew]

In principle, the additional slashed should not result in broken URL's. Compare here with here. Maybe this is related to another problem.

[fietew]

If I execute https://github.com/TWOEARS/examples/blob/master/qoe_coloration/colorationWfsCircularCenter.m I get the error:

Undefined variable "xml" or class "xml.dbGetFile".

Error in readHumanLabels (line 32)
labelFile = xml.dbGetFile(labelFile);

Error in estimateColoration (line 30)
    humanLabels = readHumanLabels(humanLabelFiles{ii});

Error in colorationWfsCircularCenter (line 34)
prediction = estimateColoration(sim, sourceFiles, sourceTypes, humanLabelFiles);

indicating that the code is still using the old database API. After fixing this, I get warnings like this:

Warning: Download failed
(url=https://dev.qu.tu-berlin.de/projects/twoears-getdata/repository/raw//experiments/2015-10-01_wfs_coloration/human_label_coloration_wfs_circular_center_noise.csv),
trying alternative database... 
> In db.downloadFile (line 43)
  In db.getFile (line 85)
  In readHumanLabels (line 32)
  In estimateColoration (line 30)
  In colorationWfsCircularCenter (line 34) 

We have set the new database (twoears-getdata) as the default database. The files, which are requested by the script do not exist anymore. However, as the API tries the alternative (old) database the download still works. The database documentation also states that the files for this coloration test have been renamed.

[hagenw]

Ah, this makes sense. I will fix this and I guess this pull request is then not really needed.
@chris-hld: or can you provide another example where the download failed from the new database, where the data should be there?

[chris-hld]

You are right, I was able to break it down to other issues in the end. I think the double slash is still not ideal, that's why I kept this PR open. It's somehow misleading while debugging. We could also just remove the last slash in the specified databases, since it is added anyhow.