The SQL readings of a text that may hold MJ template tags. Plain SQL has one view, the text
itself. A template has one view per branch variant (QueryTemplate.BranchVariants): each
renders the branches the variant takes, a {% for %} body once, and every {{ }} hole as a
placeholder of the kind its filter writes (a string literal, a number, a bare identifier, an
IN list), the way the legacy placeholder substitution did. A {{query:"…"}} composition renders
as a derived table, so it is never read as a table name. The base variant comes first.
Variants that only drop a block (an {% if %} with no {% else %} toggled off) add no SQL and
are not rendered; at most MAX_VARIANT_RENDERS others are.
The SQL readings of a text that may hold MJ template tags. Plain SQL has one view, the text itself. A template has one view per branch variant (
QueryTemplate.BranchVariants): each renders the branches the variant takes, a{% for %}body once, and every{{ }}hole as a placeholder of the kind its filter writes (a string literal, a number, a bare identifier, an IN list), the way the legacy placeholder substitution did. A{{query:"…"}}composition renders as a derived table, so it is never read as a table name. The base variant comes first.Variants that only drop a block (an
{% if %}with no{% else %}toggled off) add no SQL and are not rendered; at most MAX_VARIANT_RENDERS others are.