I've been investigating how to obtain an analysis report from the SDLSystem database behind GroupShare 2015.

I see the table (proj.AnalysisStatistics) which stores the analysis, and I can also see that there's a sproc (proj.GetProjectAnalysisStatistics) which will obtain the analysis by project ID. However, this raises some problems:

• Problem 1: The SDLSystem DB stores the analysis by LanguageFile. This means that cross-file repetitions will not be considered in cases where a project contains multiple files. In such cases we would have to collate the statistics ourselves, right?

• Not all language files have analysis statistics.
  Possible explanation: This may not be fact, but my impression is that a LanguageFile is an instance of a ProjectFile. Where a ProjectFile represents a file which has been added to a project, a LanguageFile represents that file at one phase of the project.
  Therefore, it may be the case that a LanguageFile does not have any analysis statistics until it has been worked upon, though I can't say this with any certainty.
  Indeed, if I query the GetProjectAnalysisStatistics sproc for a brand new project then it returns no statistics - possibly because the file has not yet been worked upon.

• The fields in the SDLSystem.proj.AnalysisStatistics table do not appear to match the analysis statistics which Trados makes available.
  
• When comparing the report from several projects (obtained from Trados) against corresponding entries in the proj.AnalysisStatistics table it was found that Total Words was not always the same. In both cases where I found this, the projects had two files so it could well be a case of cross-file repetitions not being counted.
  
• There's another DB table, proj.AnalysisBand which lists the fuzzy match bands, but I can't see where this is referenced.

Does anyone have any experience of obtaining analysis results direct from the GroupShare DB (SDLSystem)?

Can anyone suggest how to obtain the results for each fuzzy match band, the total (cross-file) repetitions, the perfect matches, context matches and exact matches by project?