File Exists and Exception Handling in U-Sql - u-sql

Question

Two questions
How to check file exists or not before EXTRACT?
we have scenario where new inputs file is generated every day for catalog data. we need to merge new input with d-1 file. before merge we what to make sure that new input file exists at source location
does u-sql supports try...catch block?

Answer 1

Regarding checking if a file exists. We recently released a compile-time IF statement that indeed can check for partition existence (and other objects such as files and tables are on the roadmap).
Once that feature is released (still one or two refreshs out at the time of this answer) it may look something like (syntax subject to change):
IF FILE.EXISTS("/mydir/myfile.csv") THEN
#data = EXTRACT ... FROM "/mydir/myfile.csv" USING ...;
...
#jobstate = SELECT * FROM (VALUES("job completed")) AS T(status);
ELSE
#jobstate = SELECT * FROM (VALUES("file not ready. Job not executed.")) AS T(status);
END;
OUTPUT #jobstate TO "/jobs/myjobstate.csv" USING Outputters.Csv();
You will be able to provide the name as a parameter as well. Please let me know if that will work for your scenario.
An other alternative is to use the file set syntax, especially if you want to use a dynamic value to determine the process. That would simply create an empty rowset:
#data = EXTRACT ..., date DateTime
FROM "/mydir/{date:yyyy}/{date:MM}/{date:dd}/data.csv"
USING ...;
#data = SELECT * FROM #data WHERE date == DateTime.Now.AddDays(-1);
... // continue processing #data that is empty if yesterday's file is not yet there
Having said that, you may want to check of your job orchestration framework (such as ADF) may be a better place to check for existence before submitting the job in the first place.
As to the try catch block: U-SQL itself is a script-level optimizable, declarative language where the plan gets generated and optimized at runtime over the whole script. Thus providing a dynamic TRY-CATCH is currently not available, since it would severely impact the ability to optimize the script (e.g., you cannot move predicates or column pruning outside of a try-catch block). Also TRY/CATCH can lead to some very hard to understand and debug code, especially if it is used to mimic procedural workflows in an otherwise declarative environment.
However, you can use try/catch inside your C# functions without problems if you need to catch C# runtime errors.

Answer 2

FILE.EXISTS() always returns True when executed locally. However, it works when executing against Azure Data Lake.
Tried MSDN example and the following returns True, True
DECLARE #filepath_good = "/Samples/Data/SearchLog.tsv";
DECLARE #filepath_bad = "/Samples/Data/zzz.tsv";
#result =
SELECT FILE.EXISTS(#filepath_good) AS exists_good,
FILE.EXISTS(#filepath_bad) AS exists_bad
FROM (VALUES (1)) AS T(dummy);
OUTPUT #result
TO "/Output/FileExists.txt"
USING Outputters.Csv();
I have Microsoft Azure Data Lake Tools for Visual Studio version 2.2.5000.0