Hive, mail # user - hive map field question

Re: hive map field question

Mark Grover 2011-11-23, 16:01

Hi Chalcy,
You could write a UDF that takes in a Map object and returns you the string you need.

Alternatively, you could use Hive Transform. Details here:https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Transform

When a map gets passed to your custom transform script, you could read it as a serialized string.

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Wednesday, November 23, 2011 10:48:07 AM
Subject: hive map field question
Hello HiveUsers,

I have a need to convert a map<string,string> field to string field and vice versa in a hive table. I could not do cast.

I created two external tables with one has string and another map. I can join both to get what I want, takes a long time.

Any ideas of how it can be done efficiently ?

Thanks,

Chalcy

RE: hive map field question

Chalcy Raja 2011-11-23, 16:14

Thanks, Mark. I'll try that.

Also, is there a way to list all the keys and values in a map field?

Thanks,
Chalcy

-----Original Message-----
From: Mark Grover [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, November 23, 2011 11:01 AM
To: [EMAIL PROTECTED]
Subject: Re: hive map field question

Hi Chalcy,
You could write a UDF that takes in a Map object and returns you the string you need.

Alternatively, you could use Hive Transform. Details here:https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Transform

When a map gets passed to your custom transform script, you could read it as a serialized string.

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Wednesday, November 23, 2011 10:48:07 AM
Subject: hive map field question
Hello HiveUsers,

I have a need to convert a map<string,string> field to string field and vice versa in a hive table. I could not do cast.

I created two external tables with one has string and another map. I can join both to get what I want, takes a long time.

Any ideas of how it can be done efficiently ?

Thanks,

Chalcy

Re: hive map field question

Mark Grover 2011-11-23, 16:25

Chalcy,
I don't know of an existing UDF that lists all keys and values of a map. You may want to check ql/src/java/org/apache/hadoop/hive/ql/udf in Hive source to double check.
But, as you might have guessed, it'd be pretty straightforward to do that in your own UDF:-)

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]>
To: "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
Sent: Wednesday, November 23, 2011 11:14:09 AM
Subject: RE: hive map field question

Thanks, Mark. I'll try that.

Also, is there a way to list all the keys and values in a map field?

Thanks,
Chalcy

-----Original Message-----
From: Mark Grover [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, November 23, 2011 11:01 AM
To: [EMAIL PROTECTED]
Subject: Re: hive map field question

Hi Chalcy,
You could write a UDF that takes in a Map object and returns you the string you need.

Alternatively, you could use Hive Transform. Details here:https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Transform

When a map gets passed to your custom transform script, you could read it as a serialized string.

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Wednesday, November 23, 2011 10:48:07 AM
Subject: hive map field question
Hello HiveUsers,

I have a need to convert a map<string,string> field to string field and vice versa in a hive table. I could not do cast.

I created two external tables with one has string and another map. I can join both to get what I want, takes a long time.

Any ideas of how it can be done efficiently ?

Thanks,

Chalcy

Re: hive map field question

Bejoy Ks 2011-11-23, 18:04

Hi Chalcy
Support for splitting a column of map data type into key and values is available on the hive latest trunk. You can use a explode() UDF to achieve the same, which used to support only arrays previously.

http://svn.apache.org/viewvc/hive/trunk/ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDTFExplode.java?view=markup

You need to patch your hive source installation to get this functionality running. Or as a quick fix (not highly recommended) take the source code include the same as a Custom UDF for your temporary usage. You can utilize the explode functionality as follows combining the same with Lateral Views.
Sample DDL
CREATE TABLE test_map
(
clmn1 STRING,
clmn2 STRING,
clmnMap MAP<STRING,STRING>
)

Sample DML - to extraxt key values from map
SELECT clmn1,clmn2,clmnMap_key,clmnMap_value FROM test_map
LATERAL VIEW explode(clmnMap) cl AS clmnMap_key,clmnMap_value;

Hope it helps!...

Regards
Bejoy.K.S

________________________________
From: Mark Grover <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Wednesday, November 23, 2011 9:55 PM
Subject: Re: hive map field question

Chalcy,
I don't know of an existing UDF that lists all keys and values of a map. You may want to check ql/src/java/org/apache/hadoop/hive/ql/udf in Hive source to double check.
But, as you might have guessed, it'd be pretty straightforward to do that in your own UDF:-)

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]>
To: "[EMAIL PROTECTED]" <[EMAIL PROTECTED]>
Sent: Wednesday, November 23, 2011 11:14:09 AM
Subject: RE: hive map field question

Thanks, Mark. I'll try that.

Also, is there a way to list all the keys and values in a map field?

Thanks,
Chalcy

---From: Mark Grover [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, November 23, 2011 11:01 AM
To: [EMAIL PROTECTED]
Subject: Re: hive map field question

Hi Chalcy,
You could write a UDF that takes in a Map object and returns you the string you need.

Alternatively, you could use Hive Transform. Details here:https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Transform

When a map gets passed to your custom transform script, you could read it as a serialized string.

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Wednesday, November 23, 2011 10:48:07 AM
Subject: hive map field question
Hello HiveUsers,

I have a need to convert a map<string,string> field to string field and vice versa in a hive table. I could not do cast.

I created two external tables with one has string and another map. I can join both to get what I want, takes a long time.

Any ideas of how it can be done efficiently ?

Thanks,

Chalcy

RE: hive map field question

Chalcy Raja 2011-11-23, 18:19

Wow, this is great news. I read about explode and wished map had that.

Thanks so much,
Chalcy

From: Bejoy Ks [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, November 23, 2011 1:05 PM
To: [EMAIL PROTECTED]
Subject: Re: hive map field question

Hi Chalcy
   Support for splitting a column of map data type into key and values is available on the hive latest trunk. You can use a explode() UDF to achieve the same, which used to support only arrays previously.
http://svn.apache.org/viewvc/hive/trunk/ql/src/java/org/apache/hadoop/hive/ql/udf/generic/GenericUDTFExplode.java?view=markup

You need to patch your hive source installation to get this functionality running. Or as a quick fix (not highly recommended) take the source code include the same as a Custom UDF for your temporary usage. You can utilize the explode functionality as follows combining the same with Lateral Views.

Sample DDL
CREATE TABLE test_map
(
  clmn1 STRING,
  clmn2 STRING,
  clmnMap MAP<STRING,STRING>
)

Sample DML - to extraxt key values from map
SELECT clmn1,clmn2,clmnMap_key,clmnMap_value FROM test_map
LATERAL VIEW explode(clmnMap) cl AS clmnMap_key,clmnMap_value;

Hope it helps!...

Regards
Bejoy.K.S

________________________________
From: Mark Grover <[EMAIL PROTECTED]>
To: [EMAIL PROTECTED]
Sent: Wednesday, November 23, 2011 9:55 PM
Subject: Re: hive map field question

Chalcy,
I don't know of an existing UDF that lists all keys and values of a map. You may want to check ql/src/java/org/apache/hadoop/hive/ql/udf in Hive source to double check.
But, as you might have guessed, it'd be pretty straightforward to do that in your own UDF:-)

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>>
To: "[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>" <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>>
Sent: Wednesday, November 23, 2011 11:14:09 AM
Subject: RE: hive map field question

Thanks, Mark. I'll try that.

Also, is there a way to list all the keys and values in a map field?

Thanks,
Chalcy

-----Original Message-----
From: Mark Grover [mailto:[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>]
Sent: Wednesday, November 23, 2011 11:01 AM
To: [EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>
Subject: Re: hive map field question

Hi Chalcy,
You could write a UDF that takes in a Map object and returns you the string you need.

Alternatively, you could use Hive Transform. Details here:https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Transform

When a map gets passed to your custom transform script, you could read it as a serialized string.

Mark

----- Original Message -----
From: "Chalcy Raja" <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>>
To: [EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>
Sent: Wednesday, November 23, 2011 10:48:07 AM
Subject: hive map field question
Hello HiveUsers,

I have a need to convert a map<string,string> field to string field and vice versa in a hive table. I could not do cast.

I created two external tables with one has string and another map. I can join both to get what I want, takes a long time.

Any ideas of how it can be done efficiently ?

Thanks,

Chalcy