Tuesday, May 23, 2017

Code Snippet: merging files by writing chunks

Do you have multiple source files that have to be merged so you can import them as a single one? This requirement is quite common, especially in automated data loads.

For example, our ERP system is exporting two files for Balance Sheet and Profit & Loss. We want to import them as a single file under the same POV. Merging the source files is the solution.

The approach taken is to merge files by writing chunks into target file. In this way, we avoid memory issues when having large source files.

Let's have a look!

Merging a list of files by writing chunks


'''
 Snippet:       Merge a list of files
 Author:        Francisco Amores
 Date:          23/05/2016
 Blog:          http://fishingwithfdmee.blogspot.com
 
 Notes:         This snippet can be pasted in any event script.
                Content of fdmContext object will be logged in the
                FDMEE process log (...\outbox\logs\)
                
 Instructions:  Set log level (global or application settings) to > 4 
 Hints:         Use this snippet to merge multiple single files into
                ont.
                It write chunks to avoid memory issues with large files
               
 FDMEE Version: 11.1.2.3 and later
 ----------------------------------------------------------------------
 Change:
 Author:
 Date:
'''

# initialize
srcFolder = r"C:\temp"
tgtFolder = r"C:\temp"
listSrcFilename = ["file1.txt", "file2.txt", "file3.txt"]
tgtFilename = "merge.txt"

# import section
import os
import shutil

try:
    # Open Target File in write mode
    tgtFilepath = os.path.join(tgtFolder, tgtFilename)
    tgtFile = open(tgtFilepath, "w")
    # Log
    fdmAPI.logInfo("File created: %s" % tgtFilepath)    
    
    # Loop source files to merge
    for srcFilename in listSrcFilename:
    
        # file path
        filepath = os.path.join(srcFolder, srcFilename)          
        # Log
        fdmAPI.logInfo("Merging file: %s" % filepath)                    
        # Open file in read mode
        srcFile = open(filepath, "r")        
        # Copy source file into target
        # 10 MB per writing chunk to avoid big file into memory
        shutil.copyfileobj(srcFile, tgtFile, 1024*1024*10)
        # Add new line char in the target file
        # to avoid issues if source file don't have end of line chars
        tgtFile.write(os.linesep)
        # Close source file
        srcFile.close()
        # Debug
        fdmAPI.logInfo("File merged: %s" % file)
        
    # Close target file
    tgtFile.close()
                    
except (IOError, OSError), err:
    raise RuntimeError("Error concatenating source files: %s", err)

Code snippets for FDMEE can be downloaded from GitHub.

Wednesday, May 17, 2017

Universal Data Adapter for SAP HANA Cloud

Some time ago I covered SAP HANA integration through the Universal Data Adapter (UDA). You can see details in the 3 parts I posted:
Now that everything is heading into the Cloud, why not playing around with SAP HANA Cloud?

SAP HANA Cloud
When I first tried to get a SAP ECC training environment, I noticed that SAP was offering nothing for free. Nowadays, things have changed a little bit. Luckily, they noticed that you need to offer some trial/training sandbox if you want people get closer to you.
For those who want to be part of the game, you can visit their Cloud site.

Why the Universal Data Adapter?
SAP HANA Cloud brings something called SAP Cloud Connector. Too complicated for me :-)
Luckily for me, I googled an easier way of extracting data from Cloud. There is something called database tunnels which allows on-premise systems to connect the HANA DB in the cloud through a secure connection. It doesn't sound quite straight forward but it didn't take too long to configure.

There are different ways of opening the tunnel. I have used the SAP Cloud Console Client which you can download from SAP for free.

Once the database tunnel is opened from the FDMEE Server(s) to the SAP HANA Cloud DB, the Universal Data Adapter can be used in the same way that we used with on-premise HANA DB.
Please, note that as I'm not using a productive cloud environment I had to open the tunnel via command line. This is fair enough to complete my POC.

My data in SAP HANA Cloud
I'm keeping this simple so I have a table in HANA Cloud with some dummy data:
Let's go through the configuration steps to bring that data into my application.

Importing data through FDMEE
As any configuration of UDA we need to:
  • Configure ODI Topology for the physical connection, logical schema and context
  • Configure FDMEE (source system, source adapter, period mapping, etc.)
ODI
Data Server needs to point to the DB tunnel:
We use the same JDBC driver as for HANA on-premise:
As usually, I create a dedicated context for this new source system. That gives me more flexibility:

FDMEE
In FDMEE, nothing different. 
We first create the source system with the context we created in ODI:
Then, add the source adapter for the table we want to extract data from:
Time now to import the table definition, classify columns and generate the template package in ODI:
As you can see above, FDMEE could reverse the HANA Cloud table so I can now assign the columns to my dimensions and regenerate the ODI scenario:

I'm not going to show how to create a location and data load rule as I assume you are familiar with that process.

Final step is to run our data load rule and see how data is pulled from the SAP cloud and loaded into HFM on-premise app through FDMEE :-)
I'm going to leave it here for today. As you can see, Universal Data Adapter provides a simple and transparent way of connecting our on-premise system with heterogeneous source systems, including SAP HANA Cloud!

Cheers

Wednesday, April 26, 2017

FDMEE and PBJ, together hand in hand

Do you know Jason Jones? I guess you do but in case you don't, I'm sure you may have been playing around with any of his developments.

Personally, I've been following Jason since years. I remember what I thought when I went to one his presentations in Kscope: "This guy really knows what he says and has put lot of effort helping the EPM community. Definitely an EPM Rock star."

One day, I found something quite interesting in his blog: PBJ. I thought it could be very useful to improve and simplify something that I had already built using a different solution. Why not then using something he was offering to the community as open-source? It was good to me and also good to him. I guess that seeing something you've built is useful for others, must make you proud.
When I told him that I was going to integrate FDMEE on-prem with PBCS using PBJ, he was very enthusiastic. The library was not fully tested so I made sure I was providing continuous feedback. Some days ago he published about our "joint venture". Now it's time for me.

FDMEE Hybrid Integrations
We have already covered Hybrid integrations in some posts.
In a few words, FDMEE on-prem PSU200+ can be used to extract/load data from Oracle EPM Cloud Services (E/PBCS, FCCS so far).

I suggest you also visit John's blog to know more about hybrid integrations in FDMEE:
PBJ - The Java Library for PBCS
REST Web Services, what's that? I  let you google and read about it. For us, REST is how EPM Cloud Services open to the external world. Oracle provides different REST APIs for the different EPM CS.

Luckily, Jason has gone one step further. He built a Java Library to use the REST API for PBCS:

PBJ is a Java library for working with the Planning and Budgeting Cloud Service (PBCS) REST API. It is open source software.

Why would we need PBJ in our solutions? Currently hybrid integrations have some missing functionality like working with metadata among others. For example, we recently built a solution in FDMEE to load exchange rates from HFM into PBCS.

FDMEE was offering seamless extracts from HFM. Rates are data in HFM but not in PBCS. They are treated as metadata. We used REST APIs for PBCS from FDMEE scripts which was working perfectly. However, we built the code using modules available in Jython 2.5.1. That gave rise to much head-scratching... Working with HTTP requests and JSON was not an easy task.
We noticed everything was much easier from Python 2.7 (Jython 2.7) but nothing we could do here as we were stick to what FDMEE can use :-(

TBH, we had a further ace up our sleeve: building our own Java library but we delayed this development for different reasons. It was then that PBJ appeared :-)

Why reinventing the wheel? PBJ is open-source and makes coding easier. We can collaborate with Jason in GIT and he is quite receptive for feedback given.

Using PBJ from FDMEE Scripts
When I first started testing it, I noticed that there were multiple JAR dependencies which had to be added to the sys.path in my FDMEE script. That was causing some conflicts with other Jars used by FDMEE so Jason came up with an uber-JAR:

uber-JAR—also known as a fat JAR or JAR with dependencies—is a JAR file that contains not only a Java program, but embeds its dependencies as well. This means that the JAR functions as an "all-in-one" distribution of the software, without needing any other Java code. (You still need a Java run-time, and an underlying operating system, of course.)

One of my concerns was the fact that FDMEE uses Java 1.6. That's usually a problem when using external Jars from FDMEE scripts. Luckily, PBJ is also built using Java 1.6 so the current versions of FDMEE and PBJ are good friends.

Before using any PBJ class we have to add the Jar to the sys.path which contains a list of strings that specifies the search path for modules:


# -------------------------------------------------------------------
# Add library path to sys.path
# -------------------------------------------------------------------
import sys
import os.path as osPath

# list of jar files
listPBJdep = ["pbj-pbcs-client-1.0.3-SNAPSHOT.jar"]

# Add jars
pbjDepPath = r"E:\PBJ\uber_jar"
# Debug
fdmAPI.logInfo("Adding PBJ dependencies from %s" % pbjDepPath)
for jar in listPBJdep:
    pbjPathJar = osPath.join(pbjDepPath, jar)
    if pbjPathJar not in sys.path:
        sys.path.append(pbjPathJar)
        fdmAPI.logDebug("PBJ dependency appended to sys path: %s" % jar)


Once the Jar file is added to the path we can import the different classes we want to use:


# -------------------------------------------------------------------
# Import section
# -------------------------------------------------------------------
from com.jasonwjones.pbcs.client.impl import PbcsConnectionImpl
from com.jasonwjones.pbcs import PbcsClientFactory
from com.jasonwjones.pbcs.client.exceptions import PbcsClientException
import time


We are now ready to connect to our PBCS instance.

Example: loading new Cost Centers into PBCS
I have created a custom script in FDMEE to keep it simple. The script is basically performing the following actions:
  1. Import PBJ Jar file
  2. Connect to PBCS
  3. Upload a CSV file with new Cost Centers
  4. Execute a Job to add new Cost Centers
Our CSV file with new metadata is simple, just three new members:
PBJ has class PbcsClientException to capture and handle exceptions. You can use this class in addition to Python's one:


try:
    # Your code...    
except PbcsClientException, exPBJ:
    fdmAPI.logInfo("Error in PBJ: %s" % exPBJ)
except Exception, exPy:
    fdmAPI.logInfo("Error: %s" % exPy)


Connecting to PBCS
We just need 4 parameters to create a PBCS connection:


# -------------------------------------------------------------------
# Credentials
# -------------------------------------------------------------------
server = "fishingwithfdmee.pbcs.em2.oraclecloud.com"
identityDomain = "fishingwithfdmee"
username = "franciscoamores@fishingwithfdmee.com"
password = "LongliveOnPrem"


Note: just working with Jason to use encrypted password instead of hard-coded one. I'll update this post soon.

Creating the PBJ Client (PbcsClient)
PBJ can be seen as a PBCS client built in Java so next step is to create the Client object:



# Create client
clientFactory = PbcsClientFactory()
fdmAPI.logInfo("PbcsClientFactory object created")
client = clientFactory.createClient(connection) # PbcsClient
fdmAPI.logInfo("PbcsClient object created")  


With the client object we can upload the file with new metadata to the PBCS Inbox/Outbox folder. This is done with uploadFile method:

 
# Upload metadata file to PBCS Inbox
csvFilepath = r"E:\FDMEE_CC\FDMEE_CostCenter.csv"
client.uploadFile(csvFilepath)
fdmAPI.logInfo("File successfully uploaded to PBCS Inbox")


The file is then uploaded to PBCS so the Job can process it:


Creating the Application object (PbcsApplication) 
Once file file is uploaded we need to create an application object to import new metadata. In my case, my PBCS application is DPEU.


# Set PBCS application
appName = "DPEU"
app = client.getApplication(appName) # PbcsApplication


Executing the Job and Checking Job Status
I have created a Job in PBCS to upload new cost centers from a CSV file (PBJ also supports zip files):

One thing you need to know about REST API is that they are called asynchronously. In other words, we need to check the job status until it is completed (or predefined timeout is reached).

So we first execute the job by calling method importMethod and then check job status with method getJobStatus. The status will be checked every 10 seconds while the job is running.

In order to check the job status we need to know the job id. This is done with getJobId method:


# Execute Job to import new metadata (Cost Center)
result = app.importMetadata("FDMEE_Import_CC") # PbcsJobLaunchResult
fdmAPI.logInfo("Result: %s " % result)
        
# Check Job Status and loop while executing (may add timeout)
jobStatus = app.getJobStatus(result.getJobId()) # PbcsJobStatus
fdmAPI.logInfo("Job status: %s " % jobStatus)
statusCode = jobStatus.getStatus()
while (statusCode == -1):
    time.sleep(10) # sleep 10 seconds
    jobStatus = app.getJobStatus(result.getJobId()) # PbcsJobStatus
    fdmAPI.logInfo("Job status: %s " % jobStatus)
    statusCode = jobStatus.getStatus()
        
# Show Message
if statusCode == 0:
    fdmAPI.showCustomMessage("New Cost Centers added!")
else:
    fdmAPI.showCustomMessage("Some errors happened! %s" % jobStatus)


Once the job is completed we can see the results in the PBCS Job console:
Job was executed with no errors. By navigating to the Cost Center dimension we can see the new hierarchy added:
I have also added some code to write debug entries in the FDMEE process log. This is always useful and can help you to find and fix issues easily:

Conclusion and Feedback
In this post, my main goal has been to show you how to use PBJ library in FDMEE. I'm sure this can be very useful to implement different requirements for hybrid integrations.

Jason did a great job and the ball is now in our court. The best way of contributing is to keep testing PBJ and provide feedback.
Let me highlight that PBJ is not his only project. There are few others that you can check in his site.

Enjoy FDMEE and PBJ together hand in hand!

Friday, April 7, 2017

Replacing source files on the fly - Playing around with XML files

The following post came up after seeing that lot of people in the FDMEE community were asking "how can we manipulate the source file and replace by new one on the fly?"
In other words, how we can replace the source file selected by end user with a new file we create during the import process... or let's be clear... how can we cheat on FDMEE? :-)

I thought it was good idea to share with you a real case study we had from a customer. Their ERP system had a built-in process which was extracting data in XML format. Hold on, but can FDMEE import XML files? Not out the box, Yes with some imagination and scripting.

The Requirement
As stated above, FDMEE does not support all kind of formats out of the box. We usually have to ask our ERP admin (or IT) to create a file in a format that FDMEE can easily read, mainly delimited files such as CSV.

But what about Web Services like SOAP or REST? they mainly return XML or JSON responses. We need to be prepared for that in case we want our FDMEE integration to consume a WS. This is quite useful in FDMEE on-premise as I guess Data Management for the Cloud will include "some kind of JSON adapter" sooner or later in order to integrate with non-Oracle web services.

And what about any other integration where source files are in another format than fixed/delimited?

Luckily I had that one... we get an XML file from our IT and we need to convert to CSV so we can import through FDMEE.

High Level Solution
The main idea is to convert the XML file selected to a CSV file that FDMEE can understand. Now the questions are Where and How?

  • Where? It makes sense that we do the conversion before the actual import happens. BefImport event script?
  • How? FDMEE will be expecting a CSV file, how do we convert the XML to CSV? There are multiple methods: Python modules, Java libraries for XPath... I will show one of them.
The XML File
Image below doesn't show the real XML (confidentiality) but a basic one:
As you can see data is enclosed in <data> tag an lines are enclosed in <dataRow> tags. Besides, each dimension has a different tag. 
As an extra for this post, just to let you know that I usually use Notepad++ plugin XML Tools which allows me to perform multiple operations including XPath queries:
Before we move into more details. What do you think it would happen if we try to import the XML file with no customization? 
FDMEE rejects all records in the file. What were you thinking then? That's the reason I'm blogging about this (lol)

Import Format, Location and DLR for the CSV File
In this case, our source type is File. However, I usually define instances of File when I want FDMEE admins/users to see the real source system (this is optional):
The Import Format (IF)  has been defined to import a semicolon delimited file having numeric data only (you can use any delimiter):
I'm going to keep it simple. One-to-one mapping between XML dimension tags and HFM dimensions:
The Data Load Rule is using the IF we just defined. As you may know, we can have one location with multiple DLRs using different IFs when source is File.

BefImport Event Script
The conversion will done in the BefImport event script which is triggered before FDMEE imports the file the end-user selected when running the DLR.

We can split this script into two main steps:
  1. Create the new CSV file in the location's inbox folder
  2. Update database tables to replace the original file selected with the new one created in step 1
The final solution could be more sophisticated (create the CSV based on IF definition, parse null values, etc.). Today we will go for the simple one.

Let's dive into details.

Converting XML to CSV
There are multiple ways of converting an XML to CSV. To simplify, we could group them as:
  • Method A: parses the entire XML and convert to CSV
  • Method B: converts nodes into CSV lines as we iterate them
Method A would be good for small files. It's also quite useful if our XML structure is complex. However, if we have big files we may want to avoid loading file into memory before converting which is more efficient. Therefore, I have decided to implement Method B. Within all different options we have, I will show the event-style method using xml Python module.

Which Python modules I'm using?
  • xml module to iterate the XML nodes (iterparse)
  • csv module to create the CSV file
  • os to build the file paths
Let's import the modules:

# Import Section
try:
    from xml.etree.ElementTree import iterparse
    import csv
    import os
    fdmAPI.logInfo("Modules successfully imported")
except ImportError, err:
    fdmAPI.logFatal("Error importing libraries: " % err)

Then we need to build the different paths for the XML and CSV files. We will also create a file object for the CSV file. This object will be used to create the csv writer.
The XML file is automatically uploaded to the location's inbox folder when import begins. The CSV file will be created in the same folder.

# Get Context details
inboxDir = fdmContext["INBOXDIR"]
locName = fdmContext["LOCNAME"]
fileName = fdmContext["FILENAME"]
loadId = fdmContext["LOADID"]

# XML File
xmlFile = os.path.join(inboxDir, locName, fileName)
fdmAPI.logInfo("Source XML file: %s" % xmlFile)

# CSV file will be created in the inbox folder
csvFilename = fileName.replace(".xml", ".csv")
csvFilepath = os.path.join(inboxDir, locName, csvFilename)

# To avoid blank lines in between lines: csv file 
# must be opened with the "b" flag on platforms where 
# that makes a difference (like windows)
csvFile = open(csvFilepath, "wb")
fdmAPI.logInfo("New CSV file: %s" % csvFilepath)

The writer object for the CSV file must use semicolon as delimiter so it matches with our IF definition. We have also enclosed non-numeric values in quotes to avoid issues in case you define your import format as comma delimited:

try:
    # Writer
    writer = csv.writer(csvFile, delimiter=';', quoting=csv.QUOTE_NONNUMERIC)
except Exception, err:
    fdmAPI.logDebug("Error creting the writer: %s" % err)

Once the writer is ready, it's time to iterate the nodes and building our CSV. Before seeing the code, I'd like to highlight some points:
  • We just want to capture start tags so we only capture start event in iterparse
  • We can include event in the for statement for debugging purposes (we can print how the XML file is read)
  • Property tag returns the XML node name (<entity>...)
  • Property text returns the XML node text (<entity>EastSales</entity>) 
  • We know amount is the last XML tag so we will write the CSV line when it's found
  • The CSV writer generates the delimited line from list of node texts (row)
try: 
    # Iterate the XML file to build lines for CSV file
    for (event, node) in iterparse(xmlFile, events=['start']):
        
        # Ignore anything not being dimension tags
        if node.tag in ["data", "dataRow"]:            
            continue

        # For other nodes, get node value based on tag
        if node.tag == "entity":
            entity = node.text
        elif node.tag == "account":
            account = node.text
        elif node.tag == "icp":
            icp = node.text
        elif node.tag == "custom1":
            c1 = node.text
        elif node.tag == "custom2":
            c2 = node.text
        elif node.tag == "custom3":
            c3 = node.text
        elif node.tag == "custom4":
            c4 = node.text        
        elif node.tag == "amount":
            amount = node.text 
        
        # Build CSV row as a list (only when amount is reached)
        if node.tag == "amount":
            row = [entity, account, icp, c1, c2, c3, c4, amount] 
            fdmAPI.logInfo("Row parsed: %s" % ";".join(row))        
            # Output a data row
            writer.writerow(row)
        
except Exception, err:
    fdmAPI.logDebug("Error parsing the XML file: %s" % err)

The result of this step is the CSV file created in the same folder as the XML one:
If we open the file, we can see the 3 lines generated from the 3 XML dataRows:
Cool, first challenged completed. Now we need to make FDMEE to import the new file. Let's move forward.

Replacing the Source File on the fly
FDMEE stores the file name to be imported in several tables. It took to me some time and several tests to get which tables I had to update. Finally I got them:
  • AIF_PROCESS_DETAILS: to show the new file name in Process Details page
  • AIF_BAL_RULE_LOADS: to set the new file name for the current process
  • AIF_PROCESS_PERIODS: the file name is also used in the table where FDMEE stores periods processed by the current process
To update the tables we need 2 parameters: CSV file name and current Load Id (Process Id)

# ********************************************************************
# Replace source file in FDMEE tables
# ********************************************************************

# Table AIF_PROCESS_DETAILS
sql = "UPDATE AIF_PROCESS_DETAILS SET ENTITY_NAME = ? WHERE PROCESS_ID = ?"
params = [csvFilename, loadId]
fdmAPI.executeDML(sql, params, True)

# Table AIF_BAL_RULE_LOADS
sql = "UPDATE AIF_BAL_RULE_LOADS SET FILE_NAME_STATIC = ? WHERE LOADID = ?"
params = [csvFilename, loadId]
fdmAPI.executeDML(sql, params, True)

# Table AIF_PROCESS_PERIODS
sql = "UPDATE AIF_PROCESS_PERIODS SET IMP_ENTITY_NAME = ? WHERE PROCESS_ID = ?"
params = [csvFilename, loadId]
fdmAPI.executeDML(sql, params, True)

Let's have a look to the tables after they have been updated:
  • AIF_BAL_RULE_LOADS
  •  AIF_PROCESS_DETAILS
  •  AIF_PROCESS_PERIODS
At this point, FDMEE doesn't know anything about the original XML file. Maybe some references in the process log, but nothing important.

Let's give a try!
Ready to go. FDMEE user selects the XML file when running the DLR:
Import is happening... and... data imported! XML with 3 dataRows = 3 lines imported
Process details show the new file (although it's not mandatory to change it if you don't want to)

I'm going to leave it here for today. Processing XML files can be something very useful, not only when we have to import data but in other scenarios. For example, I'm sure some of you had some solutions in mind where the Intersection Check Report (FDMEE generates an XML file which is converted to PDF) had to be processed...

I hope you enjoy this post and find it useful for your current or future requirements.

Have a good weekend!

Wednesday, March 29, 2017

BBT for FDMEE #1 - Target Accounts as Source in Mappings

Hola!
Working for customers, preparing training, conferences and the most important one, Francisco Jr running around, have been keeping me busy during the last months.

One of the presentations I've been working on is "Black Belt Techniques for FDMEE" (aka BBT for FDMEE). I thought it was interesting for people to know how to meet requirements of different complexity with some techniques which of course, aren't in the books :-)

Although I can't go into too much detail (I don't want to spoil the presentation), this is a foretaste of what you will enjoy if attending to Kscope17.

The Requirement
As you know, FDMEE can pull data from very heterogeneous source systems. Once data has been extracted, it has to be mapped into our target system (let's say HFM). Usually, people responsible of maintaining mappings (aka mappers) are more familiar with target system rather than source. 
This is not always the case but it's a common scenario when financial departments are split. How often do you hear "Not sure about this, we need to ask our ERP guy..."?

Another common scenario is that ICP/Custom dimensions mappings use source ERP account as a driver either importing source account into source ICP/Custom dimensions or using Multi-dim/Conditional maps. 

Have you have you ever asked to the mapper: Would it be easier for you if you could use the HFM account to define ICP/Custom dimension mappings rather than source account?

In my case, I always do. And what I found is that if they can define mapping using the target HFM account, maintenance tasks are much simpler and the number of mapping rules is highly reduced.

Of course, the immediate question is: Can we do that? Yes we can. How? 

Lookup dimensions as a Bridge, that's the answer
Lookup dimensions can be used in FDMEE for different purposes. How can they help us to meet our requirement?
  • They don't have an impact on target application
  • We can define a #SQL mapping to copy our target values into other source dimension values including the lookup dimension itself
  • We can define the order in which the lookup dimension mapped
Have a look at this: Source Account > Target Account > Lookup > Source C1 > Target C1
Did you understand above flow? Let's put some lights on it.

Let's start defining our lookup dimension "HFM Account":
In this example, we are going to use the lookup dimension to copy the target HFM account into the source lookup. For that purpose, we need to make sure that the lookup dimension is mapped after Account dimension. As you can see above, sequence number for the lookup is 2 while Account has assigned 1.

Besides, column UD5 of TDATASEG(_T) has been assigned to HFM Account (I could have used any other like UD10 so I leave some UDx free in case we have new HFM custom dimensions). 

Copying the Target HFM Account into other/lookup dimensions
As for any other dimension, we can create maps for lookup dimensions. Our main goal is to copy a target value into other dimensions so why not using a SQL mapping?
The good thing of mapping scripts is that we can include multiple columns:
  • Set target HFM Account to "Lookup"
  • Set source HFM Account (UD5) to target Account (ACCOUNTX)
Done, our target account has been copied and it's now available as a source value in our lookup dimension.

A picture is worth than a thousand words
Let's create a multi-dimensional mapping to show how this works:
Mapping says: when source Product is "'001" and HFM Account is "Price" then target Product is "P7000_Phones"

Thanks to our lookup dimension we can use the HFM account as a source. Mapping rule is clear and easy to create. No need to change the SQL map we created, that's static.

What happens in Data Load Workbench?
  1. Source ERP account "10100" has been mapped to "Price"
  2. "Price" has been copied to source HFM Account
  3. Product has been mapped using source Product and HFM Account
At some point, I expect Oracle enhancing multi-dim maps to support target dimensions also so let's see another example where this approach is quite useful as well.

Another example: write-back to SAP using SAP accounts as Source in Explicit maps
In this case, we are extracting data from HFM in order to generate report files for SAP (write-back)
The requirement is to map SAP IFRS using SAP Accounts. 

Following our approach:
  1. Map HFM Account to SAP Account
  2. Copy SAP Account to source SAP IFRS
  3. Map SAP IFRS based on SAP Accounts
Let's see the workflow:
As you can see, we have now copied the SAP Account into another dimension rather than the lookup. That allows us to create our Explicit mappings using SAP accounts in a very easy way.

Cloud-Friendly
Once nice thing is that this solution is cloud-friendly. Data Management for the Cloud allows creating lookup dimensions and #SQL mapping scripts so you can implement it if not using my loved on-premises FDMEE.

I'm going to leave here for today. I hope you found this BBT useful.

More BTTs soon!

Tuesday, January 17, 2017

Universal Data Adapter for SAP HANA Data Source - Part 3

Time to finish this blog series for using the UDA with SAP HANA data source. In the previous parts I covered both introduction and configuration in ODI:
In this last part we will go through the configuration in FDMEE web interface, and of course, we will extract our data from HANA.

My Source Table: FDMEE_DATA_T
Just to show you how my source data looks like:
I will be extracting data from a column-based table. I haven't tested with Virtual Tables yet so my test just shows how to extract data from standard tables.

FDMEE Artifacts
A basic configuration to make the UDA working includes:
  • Source System
  • Source Adapter
  • Source Period Mappings
  • Import Format
  • Location and Data Load Rule
  • Dummy Mappings (just to make it working)

Source System
I have registered the source system with the new context I created for HANA:

Source Adapter
Once the source adapter is created for my HANA table (FDMEE_DATA_T), we need to import the table definition into FDMEE. This is actually performing a reverse in ODI so we get the data store with its columns:
We are now able to see the columns in the source adapter. Next step is to classify Amount and Year/Period (in case we have them) columns. In my case, we do have columns for Year and Period so we can filter out data when running the extract:
I will keep display names as the column names but remember that these are the column names you will see in your import format. Therefore, I'd change them if columns have technical names.
I haven't defined any source parameter for my test but this would be done in the same as any other UDA configuration.

Once the table definition is imported and parameters are created, we have to generate the template package. This step will create the ODI interface and package:
Now we are done with the source adapter setup.

Source Period Mappings
You need to configure calendars if you are going to filter by source period and year:
If your table or view has current period data, you will be probably fine with setting period mapping type in the data load rule to None. In my case, I just created one source period map for January-2016.

Import Format
Easy one-to-one map:
After configuring the import format, we have to generate the ODI scenario that will be executed when we run the data load rule. To generate the scenario, the source adapter needs to be configured first so the table definition is imported and the ODI interface/package are successfully generated:

Location and Data Load Rule
I have created one location with one data load rule which uses source period mappings previously defined:

Running the Data Load Rule
Voilà! Here is my HANA data extracted:

Giving a try with standard View: FDMEE_DATA_V
I have also tested the data extract from a standard HANA View. Configuration steps are the same so I'm not going to replicate them. Just to show you that it works. My view is filtering only accounts starting with 1:

In case you get errors...
It may happen that you get errors when generating the ODI Scenario. In that case, I would suggest to raise a SR with Oracle as we identified some issues when working with case sensitiveness enabled.
You may get a different error but the one below shows in the FDMEE server log that the generation failed due to fatal errors in the ODI interface:

And this completes my trip around SAP HANA world. I hope you enjoyed and helped you to avoid some headaches.