U-SQL: How to skip files from analysis based on content - azure

Question

I have a lot of files each containing a set of json objects like this:
{ "Id": "1", "Timestamp":"2017-07-20T10:43:21.8841599+02:00", "Session": { "Origin": "WebClient" }}
{ "Id": "2", "Timestamp":"2017-07-20T10:43:21.8841599+02:00", "Session": { "Origin": "WebClient" }}
{ "Id": "3", "Timestamp":"2017-07-20T10:43:21.8841599+02:00", "Session": { "Origin": "WebClient" }}
etc.
Each file containts information about a specific type of session. In this case it are sessions from a Web App, but it could also be sessions of a Desktop App. In that case the value for Origin is "DesktopClient" instead of "WebClient"
For analysis purposes say I am only interested in DesktopClient sessions.
All files representing a session are stored in Azure Blob Storage like this:
container/2017/07/20/00399076-2b88-4dbc-ba56-c7afeeb9ef77.json
container/2017/07/20/00399076-2b88-4dbc-ba56-c7afeeb9ef78.json
container/2017/07/20/00399076-2b88-4dbc-ba56-c7afeeb9ef79.json
Is it possible to skip files of which the first line already makes it clear if it is not a DesktopClient session file, like in my example? I think it would save a lot of query resources if files that I know of do not contain the right session type can be skipped since they can be quit big.
At the moment my query read the data like this:
#RawExtract = EXTRACT [RawString] string
FROM #"wasb://plancare-events-blobs#centrallogging/2017/07/20/{*}.json"
USING Extractors.Text(delimiter:'\b', quoting : false);
#ParsedJSONLines = SELECT Microsoft.Analytics.Samples.Formats.Json.JsonFunctions.JsonTuple([RawString]) AS JSONLine
FROM #RawExtract;
...
Or should I create my own version of Extractors.Text and if so, how should I do that.

Answer 1

To answer some questions that popped up in the comments to the question first:
At this point we do not provide access to the Blob Store meta data. That means that you need to express any meta data either as part of the data in the file or as part of the file name (or path).
Depending on the cost of extraction and sizes of files, you can either extract all the rows and then filter out the rows where the beginning of the row is not fitting your criteria. That will extract all files and all rows from all files, but does not need a custom extractor.
Alternatively, write a custom extractor that checks for only the files that are appropriate (that may be useful if the first solution does not give you the performance and you can determine the conditions efficiently inside the extractors). Several example extractors can be found at http://usql.io in the example directory (including an example JSON extractor).