Joins in GAMS, SQL or R

To produce the charts in (1) I needed to do a multiple join on some data. From the GAMS model in (2) we extract the the following data:

What we want to arrive in R is the following:

I used approach 1 (pure SQL) in (1), but there are other ways to attack the problem.

Approach 1: pure SQL

Writing this into a SQLite database is very simple:

The database will contain copy of all the data in the model including the symbols we are interested in.

When loading the data we need to do two things:

1. Join proctime, machine and x on (j,t)
2. GAMS only stores non-zero elements, which we need to “repair”
3. Add the end column.

The second issue can be illustrated by looking at the table machine:

We miss the records that have a zero value. We can reintroduce them as follows. First we can generate the Carthesian product of J × T using a simple SQL subquery:

Adding the machine number to this 100 row table is easy with a left join:

The left join added the machine column but the missing values are represented by NA (or NULL in SQL). This is easily fixed:

We now add the other columns and also calculate the end column:

Approach 2: GAMS + SQL

It is easy in GAMS to do the join. In addition we can use a trick to keep the zeroes:

The EPS is a placeholder for the zeroes. The GAMS parameter report looks like:

The GDX2SQLITE tool will map the EPS values into zero’s, so on the R side we see:

We need to pivot on the attribute column. Unfortunately this is not completely trivial in SQL:

Approach 3: GAMS + R

We keep the GAMS part the same:

On the R side we can read the parameter report as is and pivot using spread from the tidyr package:

References

1. Gantt Chart with R/Plotly, http://yetanothermathprogrammingconsultant.blogspot.com/2017/03/gantt-chart-with-rplotly.html
2. Playing with job shop problem ft10 (1), http://yetanothermathprogrammingconsultant.blogspot.com/2014/04/playing-with-ft10-job-shop-1.html