You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hive.apache.org by John Sichi <js...@facebook.com> on 2010/04/21 20:24:30 UTC
future of CLI
Hey, if you ever get frustrated by Hive's CLI, take a look at the sqlline integration I attached here and weigh in:
https://issues.apache.org/jira/browse/HIVE-987
sqlline as is may not be exactly what we want (it doesn't have the full mysql client feel many people may be used to), but it's a good starting point, and the BSD code can be adapted as needed. If anyone wants to take over on this and really own it, it would be a great project for giving a Hive a very user-visible and beneficial facelift. If not, I'll probably at least check in the bare minimum just to scratch my own itch, unless there's opposition.
Thanks,
JVS
RE: future of CLI
Posted by John Sichi <js...@facebook.com>.
Looks like it is working for JDBC.
0: jdbc:hive://localhost:10000/default> foo bar baz;
Error: Query returned non-zero code: 11, cause: FAILED: Parse Error: line 1:0 cannot recognize input 'foo' (state=42000,code=11)
JVS
________________________________________
From: Ning Zhang [nzhang@facebook.com]
Sent: Wednesday, April 21, 2010 11:52 AM
To: hive-user@hadoop.apache.org
Subject: Re: future of CLI
Also if we use hiveserver and ODBC conneciton, we don't pass exceptions through the connection to the client. I've not tried JDBC connection yet. But if it doesn't displace error message or exceptions it may be a deal breaker.
Thanks,
Ning
On Apr 21, 2010, at 11:44 AM, Ning Zhang wrote:
> It looks cool! One questions: is it possible to run sqlline without starting hiveserver? I've seen some build that broke hiveserver. One reason is that we don't have a unit test for the standalone version of hiveserver. If we are switching hive CLI to sqlline, we need to add a unit test for that.
>
>
> On Apr 21, 2010, at 11:24 AM, John Sichi wrote:
>
>> Hey, if you ever get frustrated by Hive's CLI, take a look at the sqlline integration I attached here and weigh in:
>>
>> https://issues.apache.org/jira/browse/HIVE-987
>>
>> sqlline as is may not be exactly what we want (it doesn't have the full mysql client feel many people may be used to), but it's a good starting point, and the BSD code can be adapted as needed. If anyone wants to take over on this and really own it, it would be a great project for giving a Hive a very user-visible and beneficial facelift. If not, I'll probably at least check in the bare minimum just to scratch my own itch, unless there's opposition.
>>
>> Thanks,
>> JVS
>
Re: future of CLI
Posted by Ning Zhang <nz...@facebook.com>.
Also if we use hiveserver and ODBC conneciton, we don't pass exceptions through the connection to the client. I've not tried JDBC connection yet. But if it doesn't displace error message or exceptions it may be a deal breaker.
Thanks,
Ning
On Apr 21, 2010, at 11:44 AM, Ning Zhang wrote:
> It looks cool! One questions: is it possible to run sqlline without starting hiveserver? I've seen some build that broke hiveserver. One reason is that we don't have a unit test for the standalone version of hiveserver. If we are switching hive CLI to sqlline, we need to add a unit test for that.
>
>
> On Apr 21, 2010, at 11:24 AM, John Sichi wrote:
>
>> Hey, if you ever get frustrated by Hive's CLI, take a look at the sqlline integration I attached here and weigh in:
>>
>> https://issues.apache.org/jira/browse/HIVE-987
>>
>> sqlline as is may not be exactly what we want (it doesn't have the full mysql client feel many people may be used to), but it's a good starting point, and the BSD code can be adapted as needed. If anyone wants to take over on this and really own it, it would be a great project for giving a Hive a very user-visible and beneficial facelift. If not, I'll probably at least check in the bare minimum just to scratch my own itch, unless there's opposition.
>>
>> Thanks,
>> JVS
>
Re: future of CLI
Posted by John Sichi <js...@facebook.com>.
Oops, I made a mistake when I was testing the JDBC embedded mode. It actually works fine already, so we're good on that.
JVS
On Apr 21, 2010, at 11:59 AM, John Sichi wrote:
> Our JDBC driver docs (well, HiveDriver Javadoc) say that you are supposed to be able to use the JDBC URL "jdbc:hive://" (no hostname) to run in embedded mode (which I assume means like Derby or hsqldb). However, when I tried this, it still tried to connect to a remote Thrift server on localhost (and failed because I didn't have one running).
>
> Does anyone know if this ever worked? It doesn't seem like it should be hard to run an in-process JDBC engine (ideally without any listening socket; not sure what Thrift supports here).
>
> As long as the Hive JDBC driver can do this, sqlline just use the correct URL. That's what we do for LucidDB (allowing it to be run as either client/server or all-in-one-process engine).
>
> JVS
>
> ________________________________________
> From: Ning Zhang [nzhang@facebook.com]
> Sent: Wednesday, April 21, 2010 11:44 AM
> To: hive-user@hadoop.apache.org
> Subject: Re: future of CLI
>
> It looks cool! One questions: is it possible to run sqlline without starting hiveserver? I've seen some build that broke hiveserver. One reason is that we don't have a unit test for the standalone version of hiveserver. If we are switching hive CLI to sqlline, we need to add a unit test for that.
>
>
> On Apr 21, 2010, at 11:24 AM, John Sichi wrote:
>
>> Hey, if you ever get frustrated by Hive's CLI, take a look at the sqlline integration I attached here and weigh in:
>>
>> https://issues.apache.org/jira/browse/HIVE-987
>>
>> sqlline as is may not be exactly what we want (it doesn't have the full mysql client feel many people may be used to), but it's a good starting point, and the BSD code can be adapted as needed. If anyone wants to take over on this and really own it, it would be a great project for giving a Hive a very user-visible and beneficial facelift. If not, I'll probably at least check in the bare minimum just to scratch my own itch, unless there's opposition.
>>
>> Thanks,
>> JVS
>
RE: future of CLI
Posted by John Sichi <js...@facebook.com>.
Our JDBC driver docs (well, HiveDriver Javadoc) say that you are supposed to be able to use the JDBC URL "jdbc:hive://" (no hostname) to run in embedded mode (which I assume means like Derby or hsqldb). However, when I tried this, it still tried to connect to a remote Thrift server on localhost (and failed because I didn't have one running).
Does anyone know if this ever worked? It doesn't seem like it should be hard to run an in-process JDBC engine (ideally without any listening socket; not sure what Thrift supports here).
As long as the Hive JDBC driver can do this, sqlline just use the correct URL. That's what we do for LucidDB (allowing it to be run as either client/server or all-in-one-process engine).
JVS
________________________________________
From: Ning Zhang [nzhang@facebook.com]
Sent: Wednesday, April 21, 2010 11:44 AM
To: hive-user@hadoop.apache.org
Subject: Re: future of CLI
It looks cool! One questions: is it possible to run sqlline without starting hiveserver? I've seen some build that broke hiveserver. One reason is that we don't have a unit test for the standalone version of hiveserver. If we are switching hive CLI to sqlline, we need to add a unit test for that.
On Apr 21, 2010, at 11:24 AM, John Sichi wrote:
> Hey, if you ever get frustrated by Hive's CLI, take a look at the sqlline integration I attached here and weigh in:
>
> https://issues.apache.org/jira/browse/HIVE-987
>
> sqlline as is may not be exactly what we want (it doesn't have the full mysql client feel many people may be used to), but it's a good starting point, and the BSD code can be adapted as needed. If anyone wants to take over on this and really own it, it would be a great project for giving a Hive a very user-visible and beneficial facelift. If not, I'll probably at least check in the bare minimum just to scratch my own itch, unless there's opposition.
>
> Thanks,
> JVS
Re: future of CLI
Posted by Ning Zhang <nz...@facebook.com>.
It looks cool! One questions: is it possible to run sqlline without starting hiveserver? I've seen some build that broke hiveserver. One reason is that we don't have a unit test for the standalone version of hiveserver. If we are switching hive CLI to sqlline, we need to add a unit test for that.
On Apr 21, 2010, at 11:24 AM, John Sichi wrote:
> Hey, if you ever get frustrated by Hive's CLI, take a look at the sqlline integration I attached here and weigh in:
>
> https://issues.apache.org/jira/browse/HIVE-987
>
> sqlline as is may not be exactly what we want (it doesn't have the full mysql client feel many people may be used to), but it's a good starting point, and the BSD code can be adapted as needed. If anyone wants to take over on this and really own it, it would be a great project for giving a Hive a very user-visible and beneficial facelift. If not, I'll probably at least check in the bare minimum just to scratch my own itch, unless there's opposition.
>
> Thanks,
> JVS