currently, purgedata(or truncate) is a non-transactional operation.
It is performed by truncating the underlying traf/hbase object.
That truncate operation cannot be undone or rolled back as
it is not protected by traf transactional layer (dtm).
'drop table' on the other hand, is a transactional o
set it?
Best regards,
Yuan
-----Original Message-
From: Anoop Sharma [mailto:anoop.sha...@esgyn.com]
Sent: Friday, January 12, 2018 12:44 PM
To: dev@trafodion.apache.org
Subject: RE: purgedata can not be put in a transaction?
currently, purgedata(or truncate) is a non-transactional operation.
hi, set transaction statements are runtime stmts and only impact that session
when issued.
They are not like a cqd (control query default) that could be set at system
level.
By default, autocommit is OFF. But sqlci and trafci set it to ON at startup by
internally issuing 'set transaction autocomm
hi
you can look at file executor/ExExeUtilCli.h/cpp.
It has code/examples on how to use columnwise or rowwise rowsets
from C code in case you want to test different interfaces.
Look for cwrs* or rwrs* in that file.
java/jdbc app will be easier to use though from a user app interface.
But there
hi,
Is this only for when a connection is made to the master executor(mxosrvr)
or does it also apply when a child process (esp, udr server, arkcmp, ...) are
started
as well?
Is this for trafci only or for other clients as well, like odbc/jdbc apps, or
sqlci
or T2/T4?
Is this for the very first
yes, show remoteprocess can be used but only from trafci
and not from a standalone jdbc program.
anoop
-Original Message-
From: Roberta Marton [mailto:roberta.mar...@esgyn.com]
Sent: Wednesday, January 24, 2018 1:46 PM
To: dev@trafodion.apache.org
Subject: RE: mxosrvr debugging
Couple o
x27;s ContextCli::beginSession fits me well.
Thanks.
Regards,
Wenjun Zhu.
-邮件原件-
发件人: Anoop Sharma [mailto:anoop.sha...@esgyn.com]
发送时间: 2018年1月24日 13:31
收件人: dev@trafodion.apache.org
主题: RE: Where is the session?
hi,
Is this only for when a connection is made to the master executor(mxosrv
hi
select uuid();
syntax was added a while back based on corresponding mysql syntax.
this syntax was being used for some testing/poc at that time.
It could be removed and changed to normal function usage.
anoop
-Original Message-
From: Liu, Ming (Ming) [mailto:ming@esgyn.cn]
Sent
change to metadata tables require version change and MD upgrade.
It needs to be done carefully and in a coordinated way as MD upgrades
are only done for major releases.
As Roberta mentioned, why is TEXT table not sufficient? What kind of
metadata changes are being added?
In addition to CmpSeabas
yes, the first call to getAtttibutes with entry value of 1 is leftover obsolete
code.
It could be removed.
thanks for finding it.
anoop
-Original Message-
From: Zhu, Wen-Jun
Sent: Thursday, March 8, 2018 2:16 AM
To: dev@trafodion.apache.org
Subject: a redundant statement?
Hi,
When I
yes, that is correct. All times are returned as local time of the system where
the master executor (mxosrvr) is running.
One can also use current_timestamp_utc to return the UTC/GMT equivalent
value of current time.
If mxosrvr is running in CA and that system's clock is set to represent local
ti
latest trafodion code that will be part of traf 2.3 will generate a 36 byte
unique id
in string format.
Apache Trafodion Conversational Interface 2.3.0
Copyright (c) 2015-2017 Apache Software Foundation
>>select unique_id() from dual;
(EXPR)
Is this string representation needed for display purpose or something else?
There is a virtual 'unparse' method defined on ItemExpr class which is the
base class for value_expression.
Calling that method will return the string representation by traversing over
the tree.
This method is used for cre
ElemDDLColDef is created in parser (sqlparser.y) as part of column_definition
production
when a column is specified.
Fields are adjusted/populated in file ElemDDLCol.cpp in method
setColumnAttribute.
My suggestion would be to go through sqlparser.y, ElemDDL*.cpp,
StmtDDLCreate.cpp(and others),
In current code, varchar datatype with no length is only allowed for the
construct:
convert(, varchar)
It is changed to true in the following parser production:
| TOK_CONVERT '(' value_expression ','
{empty_charlen_specifier_allowed =TRUE;} data_type ')'
In other cases it i
in sqlci, log with no argument will stop logging.
The doc you refer to is for trafci where 'log off' will stop logging. Not all
options
listed in that doc specific to trafci are supported in sqlci.
Also note that sqlci is not externalized. It is for internal or platform usage
only.
What change
s not a block, We do a little trick with the word `exists` as
following:
* push IF onto the stack,
* when meet `exists`, check if the peek of the stack is `IF` or not, if it is,
pop the `IF` out.
We can determine if it is in block by checking if the stack is empty.
-邮件原件-
发件人: Anoop
For each insert statement, sql engine need to:
- compile the query
- begin transaction
- send data to hbase server
- commit transaction.
For simple queries like an insert, most of the cost is in the 3rd step where
values are sent to hbase server.
This is a process hop from mxosrvr to hbase region
yes, it is true that some tests do not drop all the tables that are created as
part of that test.
This is not always intentional and at times it is because one missed cleaning
them up.
But there are some advantages of not dropping tables at the end of a test run.
- drop hbase tables take a non
to recreate it
from a stripped down version of the test script. I suspect this is true more
often than not. So I haven't found this particular argument persuasive.
Regarding speeding up HBase drop: Yes, that would be a great idea.
Dave
-Original Message-
From: Anoop Sharma
Sent
With this checkin, privs2 and udr tests started to show these failures.
And they have been consistently showing up since then.
Did these pass on committer regressions run?
-Original Message-
From: steve.var...@esgyn.com
Sent: Friday, June 1, 2018 5:38 AM
To: dev@trafodion.apache.org
Subj
dealing with udfs.
anoop
From: Yang, Yong-Feng (Andy)
Sent: Monday, June 4, 2018 8:27 PM
To: Anoop Sharma ; dev@trafodion.apache.org
Subject: 答复: Trafodion regressions failures: RE: Trafodion master rh6 Daily
Test Result - 461 - Failure
Yes, it passed the regressions run. I don’t know why it would
hi
that operation is not supported as it is not a quick, metadata only change.
Every traf table under a schema has corresponding hbase object of form:
TRAFODION...
Which means that all physical HBase objects need to be renamed/copied to the
other
name of the form TRAFODION...
In addition, al
all regression failures are related to trafodion-3101 get command change.
-Original Message-
From: steve.var...@esgyn.com
Sent: Sunday, July 22, 2018 5:02 AM
To: dev@trafodion.apache.org
Subject: Trafodion master rh6 Daily Test Result - 512 - Failure
Daily Automated Testing master rh6
once the query finishes, it is removed from rms and moved to repository
unless it is a short running query.
To see explain from repository, use the following command:
(in your example, you are missing the keyword ‘QID’).
>>explain options 'f' qid mxid123456 from repository;
LC RC
hi
if the 'when' clause of a case stmt evaluates to non-true (false or null),
it should branch to the 'else' part without evaluating the 'then' part.
Since it is not doing that, this is a bug.
You can do a showplan that shows the clauses generated for the 'case'
stmt. Check why the branch claus
Constraints on other datatype would not prevent the insertion.
I cannot see why.
2. In the constraint, 1 is important, is OK.
i.e. (int1<1) prevents the insertion, but (int<) does
not.
Thank you.
Regards,
Wenjun Zhu
-邮件原件
hi Yuan
what was the reason or motivation behind your original question
of ways to get a sort_groupby plan?
Is it for a deeper understanding of a sort groupby plan, or to do perf
comparison between hash and sort groupby, or you felt that your original
query would run better with sort groupby, or
you may also be able to get a vsbb scan with rowset operations
on a single table.
something like:
select * from t where a =
This will internally be changed to a nested join with rowset of
key values being passed to the inner table being scanned.
So ultimately a nested join with multiple key pre
29 matches
Mail list logo
