By the way, I took an insight into the PlanerImpl.validate():
public SqlNode validate(SqlNode sqlNode) throws ValidationException {
ensure(State.STATE_3_PARSED);
this.validator =
new CalciteSqlValidator(
operatorTable, createCatalogReader(), typeFactory);
this.validator.setIdentifierExpansion(true);
try {
validatedSqlNode = validator.validate(sqlNode);
} catch (RuntimeException e) {
throw new ValidationException(e);
}
state = State.STATE_4_VALIDATED;
return validatedSqlNode;
}
I noticed the state is often changed in such steps, so, if this cause a
PlanerImpl instance is always bound to one SqlNode and cannot be reused to
parse/validate another SqlNodes?
———————————————
Zhihong SHEN, Ph. D., Senior Engineer
Big Data Application Service Technology Laboratory,
Computer Network Information Center, Chinese Academy of Sciences
office phone:+86-10-58812516
mobile：+86-13671116520
On 9/15/16, 1:07 AM, "[email protected]"
<[email protected] on behalf of
[email protected]> wrote:
>You have to run planner.validate after parse, otherwise the state in
>PlannerImpl will be incorrect. You can also go into the PlannerImpl and steal
>some code if you need to circumvent those states, but I agree this is probably
>the easiest way to go about it. The alternative is just creating a parser and
>SqlToRelConverter to do the parsing and conversion yourself. PlannerImpl takes
>care of a lot of other things as well like flattening nested types, view
>expansion, etc IIRC.
>
>> On Sep 14, 2016, at 2:29 AM, bluejoe <[email protected]> wrote:
>>
>> Hi, Hyde
>>
>> I wrote code like this:
>>
>> FrameworkConfig config = Frameworks.newConfigBuilder().build();
>> Planner planner = Frameworks.getPlanner(config);
>> SqlNode node = planner.parse(sql);
>> RelRoot relRoot = planner.rel(node);
>> RelNode project = relRoot.project();
>> RexNode condition = ((Filter) ((Project)
>> project).getInput()).getCondition();
>>
>>
>> However, an exception was thrown while running 'RelRoot relRoot =
>> planner.rel(node);’:
>>
>> java.lang.IllegalArgumentException: cannot move from STATE_3_PARSED to
>> STATE_4_VALIDATED
>> at org.apache.calcite.prepare.PlannerImpl$State.from(PlannerImpl.java:318)
>> at org.apache.calcite.prepare.PlannerImpl.ensure(PlannerImpl.java:108)
>> at org.apache.calcite.prepare.PlannerImpl.rel(PlannerImpl.java:200)
>>
>> How to avoid the error?
>>
>> thanks~
>>
>>
>> ———————————————
>> Zhihong SHEN, Ph. D., Senior Engineer
>> Big Data Application Service Technology Laboratory,
>> Computer Network Information Center, Chinese Academy of Sciences
>> office phone:+86-10-58812516
>> mobile：+86-13671116520
>>
>>
>>
>>
>>
>>
>>
>>
>> 在 16/9/10 上午1:25，“Julian
>> Hyde”<[email protected] 代表
>> [email protected]> 写入:
>>
>>> If you can parse & convert to get a RelNode, then the RelNode will probably
>>> be a Project with a Filter underneath. So you can write
>>>
>>> RelNode r;
>>> RexNode condition = ((Filter) ((Project) r).getInput()).getCondition;
>>>
>>> Now, how to parse a SQL statement to a RelNode? I think I’d use
>>>
>>> String sql = …;
>>> FrameworkConfig config = ...;
>>> Planner planner = Frameworks.getPlanner(config);
>>> SqlNode node = planner.parse(sql);
>>> RelRoot relRoot = planner.rel(node);
>>> RelNode r = relRoot.project();
>>>
>>> but I’d like to hear what others consider to be the most painless way to
>>> use Calcite’s parser.
>>>
>>> Julian
>>>
>>>> On Sep 9, 2016, at 2:21 AM, 沈志宏 <[email protected]> wrote:
>>>>
>>>> hi, dear all
>>>>
>>>> is there any tool class to generate RexNodes from a SQL command?
>>>> for example,
>>>>
>>>> List<RexNode> RexNodeUtils.parse(schema， "select * from persons where name
>>>> like 'm%' and age>10")
>>>>
>>>> i am writing a solr adapter and i really really need it to help write java
>>>> tests for my query translator.
>>>>
>>>> best regards
>>
>>
