Be able to dump queries to a Table in memory

I have a situation where I query a global table to get a subset of the original values (Subset A). Then, depending on the value of the first entry on that query, I need to query Subset A in various ways creating Subset B.

In order to accomplish this currently I need to dump Subset A into a "dummy" node so the values are stored somewhere, and then I am able to query Subset A normally. It would be nice if I could do something like the following:

Table myGlobalTable = reftable("GlobalTable1");

//Option 1
Table SubsetA = query("SELECT * FROM $1 WHERE Column1 > 2",myGlobalTable);

//Option 2
Table SubsetA = nullvar;
query("SELECT * FROM $1 WHERE Column1 > 2",myGlobalTable);
dumpquery(SubsetA);

Table SubsetB = nullvar
if(SubsetA[1][2] <= 10) {
	SubsetB = query("SELECT * FROM $1 WHERE Column3 = 2", SubsetA);
} else {
	SubsetB = query("SELECT * FROM $1 WHERE Column3 = 1", SubsetA);
}

This would make it so I don't need to create a "dummy" node that I will be dumping data into, but I would be able to create and store multiple queries within the same code block.

Cameron Pluim commented · Mar 29, 2017 at 06:01 PM

That's fantastic!

0 ·

Cameron Pluim commented · May 09, 2017 at 03:31 PM

@Matt Long, I just updated to 17.1.2, and I tried using a table in memory, and it doesn't seem to work with column references.

In the attached model, there are 2 script windows. In the first I am directly querying the tables in memory, and it throws errors saying:

Reference to Col 2 could not be resolved to a valid column. Parsing halted.

In the second script window I'm cloning the tables in memory to dummy nodes, and I'm able to query from the node correctly.

tableinmemory.fsm

tableinmemory.fsm (12.9 KiB)

Matt Long Cameron Pluim commented · May 15, 2017 at 01:56 PM

@anthony.johnson

anthony.johnson ♦♦ Cameron Pluim commented · May 15, 2017 at 07:49 PM

Unfortunately Table.query() was not designed for this specific situation. Instead, the new Table.query() features were meant to allow multiple query results to be active simultaneously, but these were not yet able to be subsequently passed into Table.query() again. See the documentation for p1-p9 of Table.query():

You will usually ... pass a reference to a table node here, defining the table to query ... If this parameter defines a table, it must be a tree-bound table or a direct reference to a node. In other words, it cannot be the result Table of another call to Table.query(), unless you clone that result to a tree-bound table with Table.cloneTo()

Although this might not pertain to your case, you can nest queries in 17.1. This would solve your test case, but not necessarily the original question case.

Table result = Table.query("SELECT * FROM (SELECT * FROM (SELECT * FROM $1 WHERE [Col 3] = 1) WHERE [Col 2] = 1) WHERE [Col 1] = 1", Table("GlobalTable1"));

I do believe you could solve your original question somehow, though, with a nested query

Cameron Pluim anthony.johnson ♦♦ commented · May 15, 2017 at 09:07 PM

Thanks @anthony.johnson. I didn't see that documentation, but it does spell it out pretty clearly. Should I write this up as an idea to be implemented in the future, or are there already plans to be able to do this in the future?

Also, the nested queries look really useful, but I think I'll probably end up using clone.To to a dummy node more often than nested queries from a readability stand point (depending on the table size).

I had actually set up the second test case to see what would be faster, a single query with multiple WHEREs, or multiple queries each with a single WHERE but on tables of decreasing size. But since the multiple queries would need to be copied to their own dummy nodes it made single query significantly faster.

1 ·

Show more comments

Idea

Be able to dump queries to a Table in memory

1 Comment

Write a Comment

Things to know…

Your Opinion Counts

IDEA DETAILS

Related Ideas