[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14192455#comment-14192455
]
Benjamin Lerer commented on CASSANDRA-8147:
---
{{Relation.Type}} and {{IndexExpression.Operator}} are really similar. Should
we merge them? If yes, in which package do you think that we should put the
resulting class?
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Attachments: CASSANDRA-8147-V2.txt, CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14192560#comment-14192560
]
Tyler Hobbs commented on CASSANDRA-8147:
bq. Relation.Type and IndexExpression.Operator are really similar. Should we
merge them?
I think that's a good idea, if you're willing to do it (perhaps as a second
patch, for clarity).
bq. in which package do you think that we should put the resulting class?
I'm tempted to just use Relation.Type, but if you think that's a bad idea for
some reason, I would make an o.a.c.cql3.Operator class.
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Fix For: 2.1.2
Attachments: CASSANDRA-8147-V2.txt, CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14188794#comment-14188794
]
Tyler Hobbs commented on CASSANDRA-8147:
Git bisect shows that CASSANDRA-6782 is the cause for this.
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Attachments: CASSANDRA-8147-V2.txt, CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14187032#comment-14187032
]
Tyler Hobbs commented on CASSANDRA-8147:
Overall this looks good.
The only remaining use of {{Relation.allowsIndexQueryOn()}} is in
{{SelectStatement.processRelationEntity()}}. We could remove
{{allowsIndexQueryOn()}} (and {{CollectionType.isMap()}}) by doing something
like this in {{processRelationEntity()}}:
{code}
SecondaryIndex index = indexManager.getIndexForColumn(def.name.bytes);
if (index != null index.supportsOperator(relation.operator()))
return new boolean[]{true, def.kind ==
ColumnDefinition.Kind.CLUSTERING_COLUMN};
{code}
(It looks like we need to be able to convert {{Relation.Type}} enum values to
{{IndexExpression.Operator}} enum values, though.)
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Attachments: CASSANDRA-8147-V2.txt, CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14181135#comment-14181135
]
Benjamin Lerer commented on CASSANDRA-8147:
---
I like the {{supportsOperator}} approach of CASSANDRA-8155 but I think that we
should do the validation during the preparation phase of the select statement
and not at execution time. It seems more user friendly to me.
So personally I would kind of merge the two patches as follow:
* Use the constants instead of String options and tests from this patch
* Use the {{supportsOperator}} approach of CASSANDRA-8155 instead of
{{isValidIndexFor}} in the {{SecondaryIndex}} as I found it much nicer
* Keep some validation in the SelectStatement during the preparation phase (if
you do not like the approach of this patch we can use an other one).
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Attachments: CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14181488#comment-14181488
]
Tyler Hobbs commented on CASSANDRA-8147:
bq. So personally I would kind of merge the two patches as follow
That sounds good to me. Can you take care of making that patch and I'll
resolve 8155 as a duplicate?
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Attachments: CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14180223#comment-14180223
]
Tyler Hobbs commented on CASSANDRA-8147:
Hmm, so this makes some of the same fixes as CASSANDRA-8155, but in a different
way. Personally, I prefer putting the logic for does this index support this
operator in the index code instead of the operator code (as the 8155 patch
does). However, it would be good to include your test cases and the constants
instead of String options.
Do you want to take a look at that patch and see if you agree?
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
Attachments: CASSANDRA-8147.txt
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14178080#comment-14178080
]
Benjamin Lerer commented on CASSANDRA-8147:
---
The empty result is not correct, no matter how you look at it. There are only
two things that can happen in response to the query:
you either get the good result back or you get an error with a message telling
you what you did wrong.
I agree that querying a key and a value and on map does not make a lot of
sense, so an error message will be perfectly acceptable.
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Priority: Minor
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14178171#comment-14178171
]
Sylvain Lebresne commented on CASSANDRA-8147:
-
I don't totally agree with all this. It is correct that we currently only allow
one index per CQL column, and so one cannot index both the keys and values of a
given map, but that's not what the test does. The test only has an index on the
keys.
Regarding the {{SELECT}}, provided you do have an indexed clause (which that
example has), it's allowed to have other non-indexed clause (it will require
{{ALLOW FILTERING}} but it's used in the example too). So I'm not sure why
this doesn't work, but it should (it's worth testing on current 2.1 branch
though, maybe this has been fixed since 2.1.0).
bq. I agree that querying a key and a value and on map does not make a lot of
sense
Out of curiosity, why wouldn't that make sense?
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Priority: Minor
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14178201#comment-14178201
]
Benjamin Lerer commented on CASSANDRA-8147:
---
{quote}(it's worth testing on current 2.1 branch though, maybe this has been
fixed since 2.1.0).{quote}
I tested it on the latest 2.1
{quote}Out of curiosity, why wouldn't that make sense?{quote}
As a map can only have one value associated to a given key, using such a query
means that you want to check that the key exists and that the value is the one
you think it should be. If you select using a contains key only you will be
able to have the same information but you will also know if it is the key which
is missing or the value which is not what you expect.
That is why I think that it does not make a lot of sense and that an error
message will be fine for me if I was a user.
Now as a user it is also true that it will also give me a better sense of
robustness if the query was handled properly ;-)
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14178212#comment-14178212
]
Sylvain Lebresne commented on CASSANDRA-8147:
-
bq. As a map can only have one value associated to a given key, using such a
query means that you want to check that the key exists and that the value is
the one you think it should be.
That's not what the query means, no. Asking for maps that contains a given key
and a given value does not imply that said given value must be associated to
said given key.
Besides, even if that was what the query means, the query still make sense. It
might not be terribly useful, but it make sense, so I'd still not think
throwing an error would not be very user friendly.
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Assignee: Benjamin Lerer
Priority: Minor
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
[jira] [Commented] (CASSANDRA-8147) Secondary indexing of map keys does not work properly when mixing contains and contains_key
[
https://issues.apache.org/jira/browse/CASSANDRA-8147?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14177488#comment-14177488
]
Russ Hatch commented on CASSANDRA-8147:
---
I think there is not currently support for indexing both values and keys (ref:
http://www.datastax.com/dev/blog/cql-in-2-1)
In the example above the index is on keys only, but the select is attempting to
query a values index too. So the empty result is sort of correct, but there
should probably be a warning message to prevent confusion about this (index
does not exist or something like that).
/cc [~slebresne]
Secondary indexing of map keys does not work properly when mixing contains
and contains_key
---
Key: CASSANDRA-8147
URL: https://issues.apache.org/jira/browse/CASSANDRA-8147
Project: Cassandra
Issue Type: Bug
Reporter: Benjamin Lerer
Priority: Minor
If you have a table with a map column and an index on the map key selecting
data using a contains key and a contains will not return the expected data.
The problem can be reproduced using the following unit test:
{code}
@Test
public void testMapKeyContainsAndValueContains() throws Throwable
{
createTable(CREATE TABLE %s (account text, id int, categories
maptext,text, PRIMARY KEY (account, id)));
createIndex(CREATE INDEX ON %s(keys(categories)));
execute(INSERT INTO %s (account, id , categories) VALUES (?, ?, ?),
test, 5, map(lmn, foo));
assertRows(execute(SELECT * FROM %s WHERE account = ? AND id = ? AND
categories CONTAINS KEY ? AND categories CONTAINS ? ALLOW FILTERING, test,
5, lmn, foo), row(test, 5, map(lmn, foo)));
}
{code}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
