August 13, 2015

esProc Assists Report Development – Computations Based on Multi-datasource Joins

Multiple datasources are very common in report development. We would first join tables from different databases before performing subsequent computations, such as filtering, grouping and sorting. With virtual data source or table join, reporting tools like JasperReport and BIRT can in some degree realize these computations based on joins between datasources. But they are difficult to master.

esProc, however, can be used to make the reporting tool’s handling of this situation easier, thanks to its powerful structured data computing, support for heterogeneous datasources and integration-friendly feature. The following is to illustrate how to deal with computations based on multiple datasources joins.

MySQL database has a table - sales - holding each day’s orders of more than one sellers and in which SellerId is the ID numbers of the sellers. emp is an MSSQL table having sellers’ information, in which EId is the ID numbers of the sellers, Name is their names and Dept is the departments. We want to display data of OrderID, OrderDate, Amount, Name and Dept in the report with the condition that order dates are limited to the past N days (say 30 days) or the data should belong to certain popular departments (like Marketing and Finance).

As orderID, OrderDate and Amount exist in sales while Name and Dept exist in emp, the two tables from different databases need to be joined first; then conditional filtering will be performed. Some of the source data are as follows:

Table sales

Table emp 

esProc code for doing this 

A1=myDB1.query("select * from sales")

This line of code retrieves all records from sales of myDB1, which represents MySQL database. query function is used to execute SQL queries and can receive external parameters. A1’s result is as follows: 

A2=myDB2.query("select * from emp")

This line of code retrieves all records from emp of myDB2, which represents MSSQL database. 

A3=A1.switch(SellerId,A2:EId)

This line of code switches A1’s SellerId field to its corresponding records in A2 through the relational field EId. A3’s result is as follows (data items in blue have members of lower level): 

When there is no corresponding record for a data item in A1’s SellerId, switch function will by default retain the record this data item resides but display the record’s SellerId value as null. The effect is similar to the left join. If inner join is needed, use @ioption in the function, like A1.switch@i(SellerId,A2:EId).
        
A4=A3.select(OrderDate>=after(date(now()),days*-1)|| depts.array().pos(SellerId.Dept))
This line of code filters the result of join according to two conditions. The first one, represented by the expressionOrderDate>=after(date(now()),days*-1), is to select orders during the past N days (corresponding parameter is days); the second, represented by expression depts.array().pos(SellerId.Dept), is that the orders should belong to certain specified departments (corresponding parameter is depts)The operator “||” means the logical relationship “OR”.

now function represents the current time and date function converts it into the date. after function can represent the relative time, after("2015-01-30",-30), for example, means pushing the current time back by thirty days, i.e. 2015-01-01. With different options, the function can represent the relative time based on year, quarter, month and second.

array function converts a string into a set by the delimiter. "Marketing,Finance".array(), for instance, is equivalent to ["Marketing ","Finance"]. The function’s default separator is the comma, but we can specify other separators for it. pos function locates a member in the set, ["Marketing ","Finance"].pos("Finance"), for instance, is equivalent to 2 – or true logically. If the member doesn’t exist in the set, then null will be returned - which means false logically.

Note that SellerId.Dept represents the Dept field of the corresponding record of SellerId field. It can be seen that, after fields are switched by switch function, the table relation can be represented through object style access, which is intuitive and simple, especially when establishing the multi-table and multilayer relation.

days and depts are parameters passed from the reporting tool. If they get assigned with 30 and "Marketing,Finance" respectively, A4’s result will be as follows: 

A5=A4.new(OrderID,OrderDate,Amount,SellerId.Name:Name,SellerId.Dept:Dept)

This line of code gets fields the report needs from A4. SellerId.Name
and SellerId.Dept represent respectively Name and Dept in emp. The operator “:” means renaming. A5’s result is as follows: 

Now all data are ready for the report. Finally we just need to return A5’s two-dimensional table to the reporting tool with result A5. esProc offers the JDBC interface to be integrated with the reporting tool that will identify it as a database. See related documents for the integration solution.

Design a simple report using JasperReport, for instance. Its appearance and layout are as follows: 

Define two parameters – pdays and pdepts – in the report, corresponding to the two parameters in the esProc script. Click Preview to view the report: 

The way the reporting tool calls the esProc script is the same as that it calls the stored procedure. Save this script asafterjoin1.dfx, for instance, and it can be called by afterJoin1 $P{pdays},$P{pdepts} in JasperRreport’s SQL designer.

With the assistance of esProc, the reporting tool can tackle more complicated computations based on multi-dasource joins. To find out, for example, the top three days when each seller’s sales amount increases the most rapidly after a certain date, and to display names, dates of the three days, amount and growth rate.
esProc script: 

A1=myDB1.query("select * from sales where OrderDate>=?",beginDate)

This line of code retrieves orders after a certain date from sales, in which beginDate is the parameter passed from the reporting tool, whose value let’s assume to be “2015-01-01”. Then A1’s result is as follows: 

A2=myDB2.query("select * from emp")

This line of code retrieves all records from emp as follows: 

A3=A1.switch(SellerId,A2:EId)

This line of code switches A1’s SellerId field to its corresponding records in A2. Result is as follows: 

A4=A3.group(SellerId)

This line of code groups orders by SellerId. In the following figure, the left part is A4’s result and the right part shows two orders in detail. 

A5=A4.(~.groups(OrderDate,SellerId;sum(Amount):subtotal))

This line of code groups each SellerId’s orders by OrderDate and SellerId and summarizes the amount of each group. That is, it computes the sales amount of per seller per day. The result is as follows: 

In this line of code, “A4.()” means computing A4’s members by loop. “~” in the parentheses represents a variable of members, i.e. the record of order corresponding to a certain SellerId. “~.groups()” means applying groups function to each member. groupsfunction groups data and summarizes them simply, while group function only groups data.

A6=A5.(~.derive((subtotal-subtotal[-1])/subtotal[-1]:rate))

This line of code computes daily growth rate of the sales amount of each seller. The result is as follows: 


In this line of code, derive function is used to append a new field – rate – to each group. The arithmetic is “(sales amount of the current day – sales amount of the previous day)/ sales amount of the previous day”. We can see that subtotal[-1] is used in esProc to represent the sales amount of the previous day. This makes the computing of relative position easier.

Note that, since there is not the “sales amount of the previous day” for the first record, its growth rate is Null.

A7=A6.(~.select(#!=1))

This line of code removes the first record of each group in A6 (because its growth rate is a meaningless Null). 

select function queries records we want. “#” is the loop number and thus “#!=1” means the number is not equivalent to 1. The same effect can be achieved by delete function too, but with lower performance. That’s because the former returns only the references while the latter needs to modify the real data.

A8=A7.(~.top(-rate;3))

This line of code gets records of the top three days when the growth rate of each seller’s sales amount is the biggest. topfunction gets the top N records according to a certain field (or the expression of certain fields). A8’s result is as follows: 

A9=A8.union()

This line of code unions every group of data in A8 together to create a new two-dimensional table, as shown below: 

A10=A9.new(SellerId.Name:Name,OrderDate,subtotal,rate)

This line of code gets fields as required. Then the final result is as follows: 

result A10

This line of code returns A10’s two-dimensional table to the reporting tool. See the first example for the report design, which will be omitted here.

No comments:

Post a Comment