What do you believe would be less weird?

A more intriguing question (to me): How much care should go into devising result column names for those query writers who do not really care what the name are?

What do you believe would be less weird?

sqlite> .mode box
sqlite> select 'a' /* abc */, 'b', /*abc*/ 'c';
┌───────────────┬─────┬─────┐
│ 'a' /* abc */ │ 'b' │ 'c' │
├───────────────┼─────┼─────┤
│ a             │ b   │ c   │
└───────────────┴─────┴─────┘

1 Perhaps consistency would be less weird?

• The trailing comment on the first column is retained as part of the column name.
• The leading comment in the third column is disregarded.

2 Also, removing the single quotes from literal values (when used as an identifier) that are unnamed?

sqlite> create table t1 as select 'a' /* abc */, 'b', /*abc*/ 'c';
sqlite> .dump t1
PRAGMA foreign_keys=OFF;
BEGIN TRANSACTION;
CREATE TABLE t1("'a' /* abc */","'b'","'c'");
INSERT INTO t1 VALUES('a','b','c');
COMMIT;

The present rule appears to be: Unless its column alias is given, the result column name is whatever expression yields the column value followed by other junk up to the next meaningful token, sans trailing whitespace.

This is simple to implement and understand, and makes it easy for casual, interactive query writers to correlate what is (or might be) seen as column headers to what they wrote.

I notice that nobody has undertaken an answer to my 2nd question. I suspect the answer is obvious: No further attention is needed. Certainly, there will not be some complicated set of rules devised, implemented and documented.

Perhaps, in some distant future, if given this query, select 2+3 /* A wee arithmetic demo. And a short story. /, 5 / is the Answer */; , the header will be formatted prettily and grammar-checked.

Perhaps, in some distant future, if given this query, select 2+3 /* A wee arithmetic demo. And a short story. /, 5 / is the Answer */; , the header will be formatted prettily and grammar-checked.

Not likely ... that would break baclward compatibility.

Not likely ... that would break baclward compatibility.

Not strictly speaking because the current behavior is explicitly unreliable if "AS" is not used (either explicitly or implicitly) to name the columns.