Are synthetic names prescribed by the specification, or are they implementation defined?

[RCHowell]

The PartiQL Specification uses “_1" and “_n” in some interesting ways. In the absence of some explicit bind name, "synthetic" or "fresh" names are used. It's not clear if this is implementation defined or there are some rules to how these names are enumerated.

Ask

This behavior should either explicitly defined or clearly designated as implementation defined.

Discussion Questions

• Is there a location in the specification that says these are implementation defined?
• Do we want the synthetic name rules to be implementation defined or explicit?
• How could we clarify this scoped/global counter to clean up the specification?
• Should the names be sanitized? — what happens if "_1" is already defined .. does the counter loop a next()?
• Is there a difference between "synthetic" and "fresh" per the spec wording?

Background

In the case of the enumerated synthetic names, there seems to be a "counter" — it's not clear if this counter is scoped in any way. For example,

Notice that if 𝑣1 (resp. 𝑣3) is bound to a non-tuple value 𝑣, then it is treated as if it were the tuple {'_1':𝑣1} (resp. {'_2':𝑣3}.

SELECT VALUE TUPLEUNION(
       CASE WHEN v1 IS TUPLE THEN v1 ELSE {'_1': v1} END,
       {'a':e2 },
       CASE WHEN v3 IS TUPLE THEN v3 ELSE {'_2': v3} END
)

--  Notice that the attribute names '_1', '_2' have been invented.

References to Synthetic

Notice that PartiQL extends the .* to also operate on x bindings that are not tuples. These are converted to singleton tuples with a synthetic name. — link

Notice that the input has a non-tuple that was converted to a tuple with a synthetic attribute name _1, this is because the result of a traditional SELECT is always a container of tuples. — link

References to Fresh Variable

where 𝑣 is a fresh variable, i.e., a variable that does not already appear in the query. — link

where g is a fresh variable, i.e., a variable that is not a database name nor a variable of the query or a variable of the queries within which it is nested. — link

It’s not clear the scope of n, or where synthetic names come from. Similarly the aggregations section talks about “fresh variables” which seems related, but it doesn’t follow the “_” convention.