Multirow task execution

  • 4 February 2019
  • 17 replies
  • 1049 views
Multirow task execution
Userlevel 6
Badge +9

When executing a task for a selection of rows, the user interface first executes the context procedure for every single row to check if the user is allowed to execute the task for that row, and then executes the task for each permitted row. This can be quite time consuming, especially for web clients where the progress is also reported back to the client.

To speed up multirow task execution, a new feature has been developed to execute tasks for a selection of rows.

To enable multirow execution for a task:

  1. Add an XML-typed parameter to the task
  2. Make sure Popup for each row is switched off
  3. Set the MultiselectParameterID extended property with the name of the parameter


With multirow execution enabled for a task, when the user interface executes the task, the specified XML parameter is filled with the parameter values for all rows. Even if only one row is selected.

The parameter is only filled when the task is executed. The XML is therefore not visible in the popup or available in the default and layout logic.


Example XML parameter value:

<rows>
<row>
<param1>1</param1>
<param2>a</param2>
</row>
<row>
<param1>3</param1>
<param2>b</param2>
</row>
</rows>

Be aware that both the default and the context logic are executed only once and not for every row. The parameter values for all rows are included in the XML, even if the context logic would have disabled the execution of the task for a specific row. The user interface also does not remove duplicate parameter rows when composing the XML, so you'll need to take care of that yourself.

In the task logic, use the XML parameter value instead of the regular parameters, for example:

T-SQL

select distinct
row.value('./param1[1]', 'int'),
row.value('./param2[1]', 'varchar(10)')
from @xml.nodes('/rows/row') as t2(row)


IBM i SQL

select distinct 
r.param1,
r.param2
from xmltable ('$xml/rows/row' passing v_xml as "xml"
columns
param1 integer path './param1[1]',
param2 varchar(10) path './param2[1]'
) as r;

For information on working with SQL data on SQL Server, see this link.
More info on SQL XML programming for IBM i can be found here.
 


17 replies

Userlevel 6
Badge +4
This feature works very well in conjunction with drag-drop. When you drag multiple rows, you can have a multirow-task execute once with all the dragged rows.
Userlevel 5
Badge +3
Sounds great. But I can't quite see yet whether this would solve our issue where the process in question contains a http connector. The api we're talking to, only allows a single item to be requested every time. Which means the http connector needs to be ran for every item in the selection, which (if I read correctly) is still not possible.
Userlevel 6
Badge +9
Hi Pim, good question.



A possible solution would be to create a process flow in which you recursively store the values of the first row in process flow parameters, delete the row from the XML, and then call the HTTP connector using these parameters.



The SQL code for the task or process procedure to assign the parameters would look something like this:

code:
-- Store the values of the first row
select
@param1 = row.value('./param1[1]', 'int'),
@param2 = row.value('./param2[1]', 'varchar(10)')
from @xml_input.nodes('/rows/row[1]') as t2(row)

-- Delete the first row
set @xml_input.modify('delete /rows/row[1]')


The HTTP connector should be called while the parameters have a value.

Also make sure you use a different name for the XML input parameter of the task than the MultiselectParameterID.
Userlevel 5
Badge +3
I don't think looping in processflows ever even crossed my mind. This could be a very nice solution.

One thing though, I would personally prefer keeping a pointer (int) to the current row, this would keep the input intact.



I had to fiddle around a bit, but here's an example of using a pointer:
Userlevel 5
Badge +3
code:
declare @pointer int = 1
while @pointer < 3
begin
select row.value('(./param1[1])', 'int')
, row.value('(./param2[1])', 'varchar(10)')
from @xml.nodes('(/rows/row[sql:variable("@pointer")])') as t2(row)

set @pointer += 1
end
Userlevel 3
Badge +10
I'm trying something similar with the copy file process action to copy a batch of files by obtaining "from file" and "to file" values from a select query. I put all those rows in a table variable and then loop through these rows one by one. But this is the part that eludes me:



The HTTP connector should be called while the parameters have a value.




What happens now is that the loop finishes and ends up outputing a single set of null values. How can I insert a sql "while" loop into a process flow so that it outputs parameters set by set into the process variables?
Userlevel 3
Badge +14

Does this feature also work for reports? 

I need to make a report with an output based on selected rows in a grid.

 

Does Multirow task execution still work from version 2020.1?

Userlevel 4
Badge +3

Does this feature also work for reports?

 

Reports currently don't support this. We should think about this in a design and investigate whether report implementations are able to process these kind of structured data.

I'd suggest to park this question in a separate post/idea to think about alternatives and to investigate the need for this.

Link to the parked idea:

https://community.thinkwisesoftware.com/ideas-18/multirow-report-execution-1048#post3001

Userlevel 2
Badge +2


⚠️ In the future this feature will be configurable through the metamodel and will probably use table valued parameters instead of XML. We'll be able to automatically upgrade the model, but you'll need to change the code yourself!

 

Gradually our codebase gets cluttered more and more with your originally suggested solution. Can you get us an update on when you expect the future will arrive on this matter?

Userlevel 6
Badge +9

@Ricky This is on our backlog, but we have not yet planned a date for this.

It is, however, trivial to locate and rewrite the select (and delete) from the XML to use a table valued parameter. And from a functional point of view, it doesn't make any difference.

Userlevel 5
Badge +15


⚠️ In the future this feature will be configurable through the metamodel and will probably use table valued parameters instead of XML. We'll be able to automatically upgrade the model, but you'll need to change the code yourself!

 

Gradually our codebase gets cluttered more and more with your originally suggested solution. Can you get us an update on when you expect the future will arrive on this matter?

What we've currently have done is having everywhere the same structure of creating a temporary table, which should be replaceable with the actual solution. We’re everywhere using @multi_row as input and @record as actual table used.

On each template we apply it (almost all tasks), we do something like this:

declare @record table (project_id id not null
, part_id id not null)

insert into @record (project_id
, part_id)
select x.value('(project_id)[1]', 'id')
, x.value('(part_id)[1]', 'id')
from @multi_row.nodes('rows/row') as t(x)

I hope for us that we by doing so we only need to remove the code above and can use the @records table.

 

@Ricky This is on our backlog, but we have not yet planned a date for this.

It is, however, trivial to locate and rewrite the select (and delete) from the XML to use a table valued parameter. And from a functional point of view, it doesn't make any difference.

And how to deal with unit tests on multi rows? Currently I have a long input string (<rows><row><item>1</item></row><row><item>2</item></row></rows>), but that should be converted to some kind of other data structure allowing multiple rows at once? That is less trivial than find and replace?

Userlevel 6
Badge +9

Hi Ricky,

I hope for us that by doing so we only need to remove the code above and can use the @records table.

That is indeed how it will need to be modified. 

 

And how to deal with unit tests on multi rows?

We will automatically convert these input strings to the required data structure.

Userlevel 5
Badge +15

@Ricky This is on our backlog, but we have not yet planned a date for this.

It is, however, trivial to locate and rewrite the select (and delete) from the XML to use a table valued parameter. And from a functional point of view, it doesn't make any difference.

Hi @Jasper , can you tell a bit more about the table-valued parameters in general what the plans are? And in what period of time? As I can't find it in the ‘ideas’ backlog.

I assume the multi row solution will be replaced by that, but I think I have another case for which table valued parameters could be useful. But I'm not sure it's part of the same, or something new.

It could be useful when you could design them as ‘Domain’, so it can be used in tasks, subroutines and process flows. We currently use a lot of XML passing data in process flows, but handling XML is pretty heavy for SQL Server. It also could decrease our code base as it's no longer necessary to apply transformations from a table variable, to XML, and back from XML to a table variable.

 

Userlevel 5
Badge +15

I've created a custom validation to prevent having “Popup for each row” checked while using the multi row feature (which will result in a null value). For those who are interested;

insert into validation_msg 
(
project_id,
project_vrs_id,
validation_id,
pk_col_1,
pk_col_2,
pk_col_3,
pk_col_4,
pk_col_5,
pk_col_6,
pk_col_7,
generated,
insert_user,
insert_date_time,
update_user,
update_date_time
)
select @project_id
, @project_vrs_id
, @validation_id
, @project_id as pk_col_1
, @project_vrs_id as pk_col_2
, t.task_id as pk_col_3
, null as pk_col_4
, null as pk_col_5
, null as pk_col_6
, null as pk_col_7
, 1
, dbo.tsf_user()
, sysdatetime()
, dbo.tsf_user()
, sysdatetime()
from task t
where t.project_id = @project_id
and t.project_vrs_id = @project_vrs_id
and t.popup_for_each_row = 1
and exists (
select 1
from task_parmtr tp
where tp.project_id = t.project_id
and tp.project_vrs_id = t.project_vrs_id
and tp.task_id = t.task_id
and tp.task_parmtr_id = (
select top 1 epo.property_value
from extended_property_overview epo
where epo.project_id = t.project_id
and epo.project_vrs_id = t.project_vrs_id
and epo.extended_property_id = 'MultiselectParameterID'
and epo.runtime_configuration_id = ( -- default
select rc.runtime_configuration_id
from runtime_configuration rc
where rc.project_id = t.project_id
and rc.project_vrs_id = t.project_vrs_id
and rc.default_configuration = 1
)
)
)

Use at own risk 😉.

 

Badge +1

Hi @Jasper ,

The multirow variable isn’t filled for me.

I followed the steps you described.

When I try to view the content of the variable with tsf_send_message it’s just empty for me.

Also the task still seems to be executed multiple times.

 

Can you help me with this?

 

Regards,

Peter

Userlevel 6
Badge +9

Hi pvleeuwen, 

Can you please file a ticket in TCP for this? Thanks.

Userlevel 4

I have created a dynamic control_procedure which automatically generates a @xml_table for each task that has a parameter which is the same as the property_value of MultiselectParameterID.

The goal of this procedure is that we don't have to retrieve the XML manually for all the XML variables and we have the same xml_table for each task procedure, also when we add a new parameter to the table task, this will be added automatically. This way you can use the table @xml_table in the template code for the task, instead of reading out the XML parameter every time you need some data.


Code for the template:

declare @xml_table table (
[comma][task_parmtr_id] [dom_id]
)
insert into @xml_table (
[comma][task_parmtr_id]
)
select
[comma]row.value('./[task_parmtr_id][1]', '[dom_id]')
from @[xml_variable].nodes('/rows/row') as c(row)

Code for the control_procedure (Assignment "SQL”, code group "TASKS”):

--Definition of the MultiSelectParameterID
declare @extended_property_id extended_property_id = 'MultiselectParameterID'

--Variable to save the value of the MultiSelectParameterID
declare @xml_variable property_value

--Table to save the task_id's which have a MultiSelectParameterID
declare @task table (task_id task_id)

--Retreive the tasks which have a task parameter, which have a MultiSelectParameterID connected.
insert into @task (task_id)
select t.task_id
from task t
where t.project_id = @project_id
and t.project_vrs_id = @project_vrs_id
and exists (select 1
from task_parmtr tp
where tp.project_id = t.project_id
and tp.project_vrs_id = t.project_vrs_id
and tp.task_id = t.task_id
and exists (select 1
from runtime_extended_property ep
join runtime_configuration rc
on rc.project_id = ep.project_id
and rc.project_vrs_id = ep.project_vrs_id
and rc.runtime_configuration_id = ep.runtime_configuration_id
and rc.default_configuration = 1
where ep.project_id = tp.project_id
and ep.project_vrs_id = tp.project_vrs_id
and ep.extended_property_id = @extended_property_id
and ep.property_value = tp.task_parmtr_id
)
)

--Retreive the value of the MultiSelectParameterID
select @xml_variable = ep.property_value
from runtime_extended_property ep
join runtime_configuration rc
on rc.project_id = ep.project_id
and rc.project_vrs_id = ep.project_vrs_id
and rc.runtime_configuration_id = ep.runtime_configuration_id
and rc.default_configuration = 1
where ep.project_id = @project_id
and ep.project_vrs_id = @project_vrs_id
and ep.extended_property_id = @extended_property_id
group by ep.property_value

--Connect the task to the template
insert into prog_object_item (
project_id
,project_vrs_id
,prog_object_id
,prog_object_item_id
,order_no
,control_proc_id
,template_id)
select @project_id
,@project_vrs_id
,concat_ws('_','task',t.task_id)
,concat_ws('_','task',t.task_id)
,5
,@control_proc_id
,@control_proc_id
from @task t

--Fill the variables.
insert into prog_object_item_parmtr (
project_id
,project_vrs_id
,prog_object_id
,prog_object_item_id
,parmtr_id
,parmtr_value
,order_no
,no_line_when_empty)
select @project_id
,@project_vrs_id
,concat_ws('_','task',t.task_id)
,concat_ws('_','task',t.task_id)
,'comma' as task_parmtr_id
,case when x.min_abs_order_no = tp.abs_order_no then ' ' else ',' end as parmtr_value
,tp.abs_order_no * 10
,0 as no_line_when_empty
from @task t
join task_parmtr tp
on tp.project_id = @project_id
and tp.project_vrs_id = @project_vrs_id
and tp.task_id = t.task_id
cross apply (select min(abs_order_no) as min_abs_order_no
from task_parmtr tpm
where tpm.project_id = tp.project_id
and tpm.project_vrs_id = tp.project_vrs_id
and tpm.task_id = tp.task_id
and tpm.task_id <> @xml_variable
) x
where tp.task_parmtr_id <> @xml_variable

union all

select @project_id
,@project_vrs_id
,concat_ws('_','task',t.task_id)
,concat_ws('_','task',t.task_id)
,'task_parmtr_id' as task_parmtr_id
,tp.task_parmtr_id as parmtr_value
,tp.abs_order_no * 10
,0 as no_line_when_empty
from @task t
join task_parmtr tp
on tp.project_id = @project_id
and tp.project_vrs_id = @project_vrs_id
and tp.task_id = t.task_id
where tp.task_parmtr_id <> @xml_variable

union all

select @project_id
,@project_vrs_id
,concat_ws('_','task',t.task_id)
,concat_ws('_','task',t.task_id)
,'dom_id' as task_parmtr_id
,tp.dom_id as parmtr_value
,tp.abs_order_no * 10
,0 as no_line_when_empty
from @task t
join task_parmtr tp
on tp.project_id = @project_id
and tp.project_vrs_id = @project_vrs_id
and tp.task_id = t.task_id
where tp.task_parmtr_id <> @xml_variable

union all

select @project_id
,@project_vrs_id
,concat_ws('_','task',t.task_id)
,concat_ws('_','task',t.task_id)
,'xml_variable' as task_parmtr_id
,@xml_variable as parmtr_value
,10
,0 as no_line_when_empty
from @task t

 

Result of a task which has XML as task parameter ID and has “order_line_id” and "shipping_order_id” as parameter in task parameter.

declare @xml_table table (
order_line_id id
,shipping_order_id id
)
insert into @xml_table (
order_line_id
,shipping_order_id
)
select
row.value('./order_line_id[1]', 'id')
,row.value('./shipping_order_id[1]', 'id')
from @xml.nodes('/rows/row') as c(row)

 

Reply