Archives of Support Questions (Read Only)

sea_breeze_929 · ‎02-08-2016

I have created a view to map to an existing HBase table as following:

create view "wanghai:obsweather"(pk  VARCHAR primary key,

"d"."stationid" VARCHAR,

"d"."temp"  Float,

...);

In Hbase shell, my query on a row/column yields the following:

hbase(main):004:0> get 'wanghai:obsweather', 'AP611_mesonet|2014-02-01 00:01:30', 'd:temp:toFloat'

COLUMN  CELL    

d:temp  timestamp=1454132157194, value=274.261108398437   

1 row(s) in 0.0210 seconds

In Phoenix sqlline.py, I got the following:

0: jdbc:phoenix:localhost:2181:/hbase-unsecur> select pk, "d"."stationid", "d"."temp" from "wanghai:obsweather"

. . . . . . . . . . . . . . . . . . . . . . .> where pk='AP611_mesonet|2014-02-01 00:01:30';

'PK','stationid','temp'

'AP611_mesonet|2014-02-01 00:01:30','AP611_mesonet','-0.015067715'

1 row selected (0.11 seconds)

Any idea for the same (row, cf:q), why I get "274.2611" from Hbase shell, but "-0.015067715" in Phoenix?

What is wrong?

Enis · ‎02-08-2016

HBase and Phoenix encode floats and some other data types differently.

This is from https://phoenix.apache.org/

The other caveat is that the way the bytes were serialized in HBase must match the way the bytes are expected to be serialized by Phoenix. For VARCHAR,CHAR, and UNSIGNED_* types, Phoenix uses the HBase Bytes utility methods to perform serialization. The CHAR type expects only single-byte characters and the UNSIGNED types expect values greater than or equal to zero.
Our composite row keys are formed by simply concatenating the values together, with a zero byte character used as a separator after a variable length type. For more information on our type system, see the Data Type.

View solution in original post

sea_breeze_929 · ‎02-09-2016

Thank you very much Enis!

Enis · ‎02-09-2016

No worries! Keep on HBase'ing.

Cloudera Community

Archives of Support Questions (Read Only)

Created Phoenix view to map existing HBase table, but got different float value on a row