I assume that this isn't all that important but would like to understand what is taking place. I had some longer CTEs that needed split because one of the intermediate tables needed to be used in another location and would be lost when the CTE completed; thus, I wrote it to an in-memory database and then continued the CTE in a second query.
dbcmd eval $sql
# split into
dbcmd eval ${sql_a}${sql_b}
Is this more work for Tcl than using two dbcmd eval such as:
dbcmd eval $sql_a
dbcmd eval $sql_b
Does Tcl first append those two SQL strings together in another memory location, when it might just pass a pointer to each eval and do less?
Unless query results are being written back to the application code, I tend to try to put as much in a single eval as possible but really have no reason for that. Is there a preferred approach?
If it matters, nearly all the queries are in a separate namespace and are not built within the script other than declaring them within {}s and using the : operator to bind parameters.
Thank you.
The difference in the amount of work to Tcl is trivial relative to the cost of even accessing an in-memory database. You should write code that both works and is as clear to you as practical.
There is no concatenation of the pieces in
dbcmd eval $sql_a
dbcmd eval $sql_b
The $sql_a and $sql_b strings should each contain complete statements (or several of them) as that is what the underlying API accepts.
You are recommended to always use {brace-quoted} literal strings for SQL in Tcl. String substitutions to generate SQL can be (and often is) a sign of critical vulnerability of the code to SQL injection attacks.