And we’re back. Sorry for the delay. Between day job, the worlds worst sinus infection, and becoming a pee-wee soccer coach (which required actually learning the rules of soccer), it’s been a busy end of summer.
Part n of our m part series on events continues. Today we’re going to talk about the Before Database Generate event. I could take the easy road by pointing out that I’ve already covered these in my post on adding seed data to a table but we’ll take it a little further.
First, let’s talk about before database generate events (and their cousin after database generate events) in general. What can we use them for? Well, here’s the list I came up with.
Now, just because you can do these things doesn’t necessarily mean you should. This event is at the MODEL level. This should give you an idea that Sybase really intended this as something you’d use for versioning, external tool interface, etc. rather than for manipulating individual objects (that’s what transforms are for) prior to generation. You can still write code to go through your objects and take whatever action you want, but think long and hard about it.
I’ve covered external integration with my seed data demo and Sybase has some good material provided on interacting with the repository, so for our example I’m going to modify objects based on the date.
I’m going to configure my model so that for any date range partitioned table, I only have to specify the starting partition, and it will built out partitions until the current date. I’ll be honest, this isn’t how I would choose to do this, in part because my sample is too simplistic to really be useful, in part because most databases now have interval partitions where you only have to tell it how long the date range for each partition is and it takes care of the rest, and in part because if this is something I need to automate in a warehousing environment, it’s usually something I prefer to build into the ETL or operations routine anyway. But, it’s still a fun example, and it gave me an excuse to play with the GetPhysicalOptionValue method.
The idea here is we’ll define a single partition at the beginning of your date range. If I wanted to make this more flexible, I’d establish what the interval was for each partition (daily, weekly, monthly, quarterly, yearly) as an extended attribute and the routine would take care of the rest using the options on the initial partition as a template. If you take it this far, let me know.
We’ll start with a custom method on the table object. Since partitions are fairly database specific, I should probably put this in the XDB file but I’m going to leave it in an XEM simply to make it easier to distribute. This particular method is based on Oracle 11g.
Here’s the code for my method. I’ll walk through it below.
sub %method%(obj) 'BuildPartitions method dim part dim part_col_code dim part_col dim col dim part_value dim start_dt dim part_cnt dim myregexp dim mymatches dim mymatch if not obj.getextendedattributetext("tablepropertiestablepartitioningclausesrangeorcompositepartitioningclausepresence") then output "table " & obj.code & " is not range partitioned" exit sub end if ' now assuming we have range partitions, is it a date? get the name of the column part_col_code = obj.getphysicaloptionvalue("<table_partitioning_clauses>/partition by range/<column_list>/<column>:1") 'now find that column object for each col in obj.columns if col.code = part_col_code then set part_col = col end if next ' and see if it is a date. if part_col.datatype = "DATE" then 'now we retrieve the partition data for the single partition we've defined part_value = obj.getphysicaloptionvalue _ ("<table_partitioning_clauses>/partition by range/<partition_list>/partition:1/values less than/<value_list>") 'now lets get the current date and build out partitions to support it set myregexp = new regexp myregexp.ignorecase = true myregexp.global = true myregexp.pattern = "\d+/\d+" set mymatches = myregexp.execute(part_value) for each mymatch in mymatches start_dt = cdate(mymatch.value) next output "Building partitions from " & start_dt & " to " & date & " on " & obj.Code else output "table " & obj.code & " is not range partitioned by date" exit sub end if ' now we know we have a table that is range partitioned by a date column ' let's add monthly partitions from start date to current date + 1 (remember ' the partition is defined by the upper bound ' set the partition count - we have one defined by default, so we'll start with 2 part_cnt = 2 do while start_dt <= dateadd("m",1,date) obj.SetPhysicalOptionValue "<table_partitioning_clauses>/partition by range/<partition_list>/partition:" & _ part_cnt & "/<partition_name>", "part" & year(start_dt) & Right("0" & month(start_dt), 2) obj.SetPhysicalOptionValue "<table_partitioning_clauses>/partition by range/<partition_list>/partition:" & _ part_cnt & "/values less than/<value_list>", "to_date('" & year(start_dt) & "/" & Right("0" & month(start_dt), 2) & "','YYYY/MM')" start_dt = dateadd("m",1,start_dt) part_cnt = part_cnt + 1 loop output part_cnt & " partitions built" end sub
Let’s break it down. Our first task is to determine if the table is range partitioned, and if it’s a date. To do this, we’ll check and extended attribute on the table, then we’re going to inspect the physical options on the table to get the partitioning column we’ve defined and determine if it’s a date.
Lines 15-19 check the partitioning and terminate the subroutine if it’s not range partitioned. We follow that immediately by using the getPhysicalOptionValue method to retrieve the name of the partitioning column. I don’t find the help file on this method terribly informative. So I’ll give you my Cliff’s Notes version of common errors.
Lines 26-33 find the column object referred to in the physical options by looping through the columns collection. If the column is not a date, the subroutine will exit.
Lines 36 and 37 retrieve the “values less than clause”, again using the getPhysicalOption method.
I’m counting on my method having the value expressing in some sort of standard date format. VBScript is pretty good at automatic date conversions, but use your best judgement here. Lines 41-50 use regular expressions to extract the date portion from the value. I’m extracting all numbers and the “/” character. If you want a good tutorial on regular expressions, you are NOT on the right blog. I know enough to be dangerous and how to use Google. Here are two sites I find handy:
Now that we’ve completed all of that work, we’ll loop through from our starting date to the current date (plus one month since we’re specifying an upper bound) in lines 66 to 75 and add partitions using the setPhysicalOptionValue method. The same warnings apply as for getPhysicalOptionValue(s). Notice that we need to change our index to build new partitions and that we start with 2 so we don’t overwrite the starter partition.
So, a lot of work and still no BeforeDatabaseGenerate event (I guess maybe I should’ve titled this one “Getting and Setting Physical Options”). Let’s add it now.
I’m going to assume you’re familiar with adding classes and events to an XEM file, we’ve covered that before. Remember it’s got to be at the MODEL level. Here’s the code:
Function %BeforeDatabaseGenerate%(model, fct) Dim myObj Dim mySel Set mySel = model.CreateSelection mySel.AddObjects ActiveModel, cls_table, false, true For each myObj in mySel.Objects With myObj Output .Name .ExecuteCustomMethod("Events2.BuildPartitions") end with Next %BeforeDatabaseGenerate% = True End Function
Nothing surprising here at this point (I hope). As usual, we’re using a selection to find all our tables, and then executing our method for each. You’ll notice that the input parameter for the method isn’t Obj for this event, since it only applies to models, it’s been replaced with the more specific Model on lines 1 and 6.
So there you have it. A few words of wisdom (or maybe just words) on what you can and should do with BeforeDatabaseGenerate events, and a brief sample. Remember, this event is probably best reserved for integration with external files or sources and version control rather than actual model manipulation. Use transforms for that. However, in a pinch, here’s another tool for the box.
Best of luck.