D4Science Infrastructure

We cannot assume here that file names always have or always miss extension (I think also it depends on the particular algorithm in case of SM integration), so the portlet should check it before eventually adding the extension to the filename. I think it's a better approach to adjust the names in the operation that handles it before presenting it to the portlet (or where it is generated, where possible).
Please note also that in case of SM integration, mimetype is only available for resources retrieved from the workspace (SMType IMAGES).

The problem is that a invalid MimeType is set. PNG image has MimeType "image/png", instead is set "png".

ResourceDescriptorTD [id=152, name=Chart focused on decimallatitude - Distribution of summed quantities over emerged lands, divided per country, 
description=Charts, 
creationDate=2015-07-01 11:17, 
creatorId=10001, 
resourceType=Generic File, 
resourceTD=InternalURITD [id=5593cc52e4b00f093ea5d524, mimeType=png, thumbnailTD=null]]

Gianpaolo, can you provide a list of expected mimetypes that we could check/fix when we import resources onto Tabular Data?
If not we could :

• proceed by evidence
• use an external library that analyses the imported file and gives us the proper mimetype (definitely more secure but not optimal because we need to move all files from the storage to the Tabular data service)

The files currently produced by StatMan are:
png
csv
gif
zip
txt
no-extension (binary)

The images produced by the algorithms are always generic Java "Image" objects. They are always saved by StatMan as PNGs on the workspace. Thus, there will be no JPG or other formats for images.

The modification asked for this ticket regards the PNGs only, as they should be visualized as images and downloaded as PNGs. These are reported as IMAGE content by StatMan (thus they are recognizable also from the category) and are the only kind of image reported by the service. Anyway, the mime-type should be correctly reported by the workspace, isn't it?

The other files are classified as FILEs by StatMan and their mime-type can be retrieved from their extension. These files always have an extension, unless they are binary files. In this case mime-type is not really important, because these are expected to be treated as generic files.

The images now work, but I see problems in other kinds of files:
String results are reported with a generic STRING name and not with their real associated key, whereas the Files object names are reported with their associated keys and not with the file name produced by the statistical manager.

E.g. the output of "Feed Forward Ann" is a "General Results.txt" file plus a binary file. The TXT should contain the string "Learning threshold=0.56" and instead contains " [STRING] : 0.56". Furthermore, the binary file accompanying this TXT should have the name introduced by the user as input, instead, as result on TabMan, it is transformed into a generic "Trained Neural Network" string. This is the key of the File object sent by the Statistical Manager but it is not the file name (which has instead the name introduced by the user).