[HBase-dev] Is there a API to get all the columns of a column family

Its not possible without getting all data in all columns; no two rows
necessarily have the same columns. But perhaps there was some
discipline in place inserting data in your case and retrieving all
data on a particular row will tell you the columns used on all rows?

St.Ack

I'll add this to the book. This comes up from time to time.

Please note there is also
https://issues.apache.org/jira/browse/HBASE-4882 which
suggest to add such feature, of course this is just for a specific row and
does not tell that all the row are the same.

Mikael.S

Hi St.Ack,

Thanks for your quick response. In my case, the tables comes from customers
so I can't modify those tables, and some of the tables are very large(for
example, one is more than 600MB) , when I scanning all the data of the
table, the program just does not respond. Do you have a suggestion about
this kind of case? And I think more and more people will encounter this
problem, is that possible in future to add the API to get all the column
names of a column family(such as HColumnDescriptor.getColumnNames() )?

Kind regards,

Hongqi

Why not? It stops at a particular region? Can you figure which one.
Is it not deployed anywhere? Grep its name in master logs to see if
you can figure what happened to it?
I don't think we'll be adding such a call; it would imply a full scan
of the whole table to find every unique column of which there could be
millions. If the table is big, it'd need to mapreduce for the result
to come back in any reasonable amount of time.

St.Ack

Sorry, St.Ack , I'm not very clear about how to mapreduce for the result,
could you explain more about the mapreduce or give me some link? Thanks
again for your kind help!


hongqi

There are lots of examples of MapReduce using HBase here...

http://hbase.apache.org/book.html#mapreduce

HBase table schema's only pre-define the table's column families, but not
the columns in it. In the data model, every row defines its own columns, so
every row potentially can contain different columns. Even if your data has
fixed columns, there is no way that hbase can determine that. The
HTableDescriptor object also is not the place for this, since it defines
the schema of the table, and column names are not in the schema. Please see
the book on the data model http://hbase.apache.org/book.html#datamodel.

Having said that, if your data has fixed columns, you can manage those
definitions externally, or store them in a meta table of your own. Or if
you want to obtain column names for a single row, you can do so by calling
Htable.get(), and obtaining the column names from the keys of
Result.getFamilyMap(). If you are sure that all the rows contain the same
columns, then, you can do something like, get the first row, get the
columns of the first row.

Thanks,
Enis