Here's a distilled version of what we're trying to do. The transformation step is a "Table Input":
SELECT DISTINCT ${SRCFIELD} FROM ${SRCTABLE}
We want to run that SQL with variables/parameters set from each line in our CSV:
SRCFIELD,SRCTABLE
carols_key,carols_table
mikes_ix,mikes_rec
their_field,their_table
In this case we'd want it to run the transformation three times, one for each data line in the CSV, to pull unique values from those fields in those tables. I'm hoping there's a simple way to do this.
I think the only difficulty is, we haven't stumbled across the right step/entry and the right settings.
Poking around in a "parent" transformation, the highest hopes we had were:
We tried chaining CSV file input to Set Variables (hoping to feed it to Transformation Executor one line at a time) but that gripes when we have more than one line from the CSV.
We tried piping CSV file input directly to Transformation Executor but that only sends TE's "static input value" to the sub-transformation.
We also explored using a job, with a Transformation object, we were very hopeful to stumble into what the "Execute every input row" applied to, but haven't figured out how to pipe data to it one row at a time.
Suggestions?
Aha!
To do this, we must create a JOB with TWO TRANSFORMATIONS. The first reads "parameters" from the CSV and the second does its duty once for each row of CSV data from the first.
In the JOB, the first transformation is set up like this:
Options/Logging/Arguments/Parameters tabs are all left as default
In the transformation itself (right click, open referenced object->transformation):
Step1: CSV file input
Step2: Copy rows to result <== that's the magic part
Back in the JOB, the second transformation is set up like so:
Options: "Execute every input row" is checked
Logging/Arguments tabs are left as default
Parameters:
Copy results to parameters, is checked
Pass parameter values to sub transformation, is checked
Parameter: SRCFIELD; Parameter to use: SRCFIELD
Parameter: SRCTABLE; Parameter to use: SRCTABLE
In the transformation itself (right click, open referenced object->transformation):
Table input "SELECT DISTINCT ${SRCFIELD} code FROM ${SRCTABLE}"
Note: "Replace variables in script" must be checked
So the first transformation gathers the "config" data from the CSV and, one-record-at-a-time, passes those values to the second transformation (since "Execute every input row" is checked).
So now with a CSV like this:
SRCTABLE,SRCFIELD
person_rec,country
person_rec,sex
application_rec,major1
application_rec,conc1
status_rec,cur_stat
We can pull distinct values for all those specific fields, and lots more. And it's easy to maintain which tables and which fields are examined.
Expanding this idea to a data-flow where the second transformation updates code fields in a datamart, isn't much of a stretch:
SRCTABLE,SRCFIELD,TARGETTABLE,TARGETFIELD
person_rec,country,dim_country,country_code
person_rec,sex,dim_sex,sex_code
application_rec,major1,dim_major,major_code
application_rec,conc1,dim_concentration,concentration_code
status_rec,cur_stat,dim_current_status,cur_stat_code
We'd need to pull unique ${TARGETTABLE}.${TARGETFIELD} values as well, use a Merge rows (diff) step, use a Filter rows step to find only the 'new' ones, and then a Execute SQL script step to update the targets.
Exciting!
Related
I have an excel with 300 rows. I need to use each of these rows as a field name in a transformation.
I was thinking of creating a job that for each row of a table sets a variable that I use afterwards on my transformation.
I tried defining a variable as the value I have in one row and the transformation works. Now I need a loop that gets value after value and redefines the variable I created then executes the transformation.
I tried to define a Job that has the following:
Start -> Transformation(ExcelFileCopyRowsToResult) -> SetVariables -> Transformation(The transf that executes using whatever the variable name is at the moment).
The problem is that the variable I defined never changes and the transformation result is always the same because of that.
Executing a transformation for each row in a result set is a standard way of doing things in PDI. You have most of it correct, but instead of setting a variable (which only happens once in the job flow), use the result rows directly.
First, configure the second transformation to Execute for each row in the Edit window.
You can then use one of two ways to pass the fields into the transformation, depending on which is easier for you:
Start the transformation with a get rows from result. This should get you one row each time. The fields will be in stream directly and can be used as such.
Pass the fields as parameters, so they can be used like variables. I use this one more often, but it takes a bit more setup.
Inside the second transformation, go to the properties and enter variable names you want in the Parameters tab.
Save the transformation.
In the job, open the transformation edit window and go to Parameters.
Click Get Parameters.
Type the field name from the first transformation under Stream Column Name for each parameter.
I am researching standard sample from Pentaho DI package: GetXMLData - Read parent children rows. It reads separately from same XML input parent rows & children rows. I need to do the same and update two different sheets of the same MS Excel Documents.
My understanding is that normal way to achieve it is to put first sequence in one transformation file with XML Output or Writer, second to the second one & at the end create job with chain from start, through 1st & 2nd transformations.
My problems are:
When I try to chain above sequences I loose content of first updated Excel sheet in the final document;
I need to have at the end just one file with either Job or Transformation without dependencies (In case of above proposed scenario I would have 1 KJB job + 2 KTR transformation files).
Questions are:
Is it possible to join 2 sequences from above sample with some wait node before starting update 2nd Excel sheet?
If above doesn't work: Is it possible to embed transformations to the job instead of referencing them from external files?
And extra question: What is better to use: Excel Output or Excel Writer?
=================
UPDATE:
Based on #AlainD proposal I have tried to put Block node in-between. Here is a result:
Looks like Block step can be an option, but somehow it doesn't work as expected with Excel Output / Writers node (or I do something wrong). What I have observed is that Pentaho tries to execute next after Block steps before Excel file is closed properly by the previous step. That leads to one of the following: I either get Excel file with one empty sheet or generated result file is malformed.
My input XML file (from Pentaho distribution) & test playground transformation are: HERE
NOTE: While playing do not forget to remove generated MS Excel files between runs.
Screenshot:
Any suggestions how to fix my transformation?
The pattern goes as follow:
read data: 1 row per children, with the parent data in one or more column
group the data : 1 row per parent, forget the children, keep the parent data. Transform and save as needed.
back from the original data, lookup each row (children) and fetch the parent in the grouped data flow.
the result is one row per children and the needed column of the transformed parent. Transform and save as needed.
It is a pattern, you may want to change the flow, and/or sort to speed up. But it will not lock, nor feed up the memory: the group by and lookup are pretty reliable.
Question 1: Yes, the step you are looking after is named Block until this (other) step finishes, or Blocking Step (untill all rows are processed).
Question 2: Yes, you can pass the rows from one transformation to an other via the job. But it would be wiser to first produce the parent sheet and, when finished, read it again in the second transformation. You can also pass the row in a sub-transformation, or use other architecture strategies...
Question 3: (Short answer) The Excel Writer appends data (new sheet or new rows) to an existing Excel file, while the Excel Output creates and feed a one sheet Excel file.
I have just started learning pentaho spoon steps and have one problem with solving one problem. I need to transform the data from xls-file and convert it do database. The problem is that my input file looks like this: table-description
And I can not find how to solve two problems:
For my next step I need to save not only the table itself (Range A8:D11), but also the date (cell A5). When I am trying to do it in pentaho with Microsoft Excel Input – Step it works only when I select A8-cell as a start row, but the date is not saved.
In Microsoft Excel Input – Step I must always select a start row in order to generate a table and use it in next steps. And I must do it manually, I mean to say that my table starts from A8-cell. In my case I can not always say for sure that the table starts from A8-cell. I know, that the start-cell is that cell, which is in A-Column and has value = “Date”. Microsoft Excel Input – Step will be first step in my kettle because I must get data and change them. That is why I think I can not use before Java Script.
I have not found the solution to these two problems and I do not know if it is possible to make it. I will be grateful for any help.
I am not sure what do you mean by converting an excel file to database but If you can convert the xls into csv and read that file then you know from which row you need to filter the data. Basically you can use a simple filter step to filter the data when it matches column name. I hope this will help.
Use two Microsoft Excel Input steps. One step reads the table (A8:D11). The other step reads the date (A5). Then merge the two streams, for example using a Join Rows (cartesian product) step
Read everything. Then use a Javascript step with two script tabs. For one of the tabs: Right-click and choose Set start script. Code : var start = 0; The other tab should be kept as a transformation script. Pseudocode: if(FieldA equals "Date") {start = 1;}. Now you will have an additional field in the stream called start. If start equals 0, then you know that your tabular data hasn't started yet, and you can filter out the row.
I'm trying to process some data and store it in a datawarehouse. For doing it, I wanted to store dimensions in one transformation and fact (only have one) in another transformation. So I can use a job for execute the first one, copy rows to result and get them into the second transformation.
In the first transformation, I read some Excel file and separate this data into some streams. It is data from a baptism, so I have one stream for the person, another one for parents, another one for sponsors, and so on... At the end of each stream, I insert data into database and return PK autogenerated (it is an id autoincrement).
In the second one, I only have Get rows from result and want to set them into a txt file (just for see it is been done correctly). The problem is that the file is created but it is empty. I suppose that if I let fields in Get rows from result empty, it gets all fields.
What am I doing wrong?
At the end what I want is to have one Copy rows to result at the end of each stream in the first transformation and get all this data in the second one.
In "Insert Pare Padrina" I return id_pare_padrina which is autogenerated, and the same with "Insert Mare Padrina" (I have more streams which I also have to include them into result). This transformation is not executed per row because I need values of other rows.
Thank you!
In order to pass the data from the first transformation to the second transformation, you need to set certain parameters like:
1. First of all, in the transformation settings of the second transformation (at the Job Level), check on the items as image below:
Copy Previous results to parameters will ensure that all the results/data in the "Copy Rows to Result" step is getting properly passed to the next level.
Execute for every input row : will execute the second transformation for every rows in the first transformation file. This is optional based on your requirement.
2. In the same transformation settings, define the "Parameters" in the Parameters tabs. Check the image below:
Here, NAME is the parameter i have defined. So when you are using the "Get rows from result", you can define these parameter names.
3. Instead of using "Get rows from result", you can alternately use "Get Variables" step to fetch all the variables coming from the previous step. All you need to do is to define the parameter names inside the ktr file (CTRL + T). (Actually i have practically implemented in that fashion and it worked for me.)
4. Since "copy rows to result" step uses heap memory, defining multiple instances of this step might exhaust the memory space quickly and your code might fall in trouble. Ideally use a single instance of this step.
But if your data interation is only one row, best option would be to use "set variables" step.
I assume you might have missed some of these sections in the job.
You can read more on copy rows to result in here.
Hope it helps :)
I have an SSIS package that runs a script component as one step in a series that transforms data from a flat file to a SQL table. The script component itself is pretty straightforward, but I had a question about its input and output columns.
The script appears to have no output columns defined (there are a number of input columns, of course.) Yet when I run the SSIS package, data comes out of this script component -- it's then used as input for a data conversion component, and from there pushed into a SQL table.
Is there a default setting I'm not aware of, where a script component with no defined output columns defaults to using the input columns? Thanks for helping me clear this up.
The OUTPUT Columns section is for defining columns that you are adding to the output after the script completes. In other words, if you are taking several values from the data flow in and based upon their values, calculating a new value to be output in a new column, then that would be defined in the script as an output. Otherwise, the buffer that is input into the script task is output out of the script task.