It might be simpler to add a .q file testcase.
Just load 2 files (say a1.q and a2.q in a hdfs directory).
Then load a new file, say foo, for the table 'T' - the contents of the file 'foo' are

a1.q
a2.q

Then, 'T' can be queried

Also, it might be simpler to add the new function 'getContentSummary' in all existing
input formats.

You can create a dummy class which all other hive input formats (other than symlinktextinputformat) extend.
In the abstract dummy class, the existing defn. can be there.

FileSystem fs = p.getFileSystem(ctx.getConf());
cs = fs.getContentSummary(p);

That waym you dont need any special checking in Utilities.java - it calls getContentSummary(),
which is implemented by all input formats that hive supports.

namit, a .q test file can not include what this jira does. From a .q file, it is very difficult to know SymlinkTextInputFormat get the input size correctly.

>>getContentSummary' in all existing input formats.
There is no guarantee that the inputformat is from Hive. It is very difficult to change all input format.

fixed a typo

+1

Committed. Thanks Yongqiang!