datafusion-cli: ignore hashbang at top of script

[stephenmw]

Is your feature request related to a problem or challenge?

I would like to create executable scripts using datafusion-cli as an interpreter.

For example, given the following tmp.sql file:

#!/usr/bin/env -S datafusion-cli -f
SELECT 1;

I would like to be able to run the following successfully:

$ chmod 755 tmp.sql
$ ./tmp.sql

Instead, I get this:

$ chmod 755 tmp.sql
$ ./tmp.sql
DataFusion CLI v40.0.0
SQL error: ParserError("Expected an SQL statement, found: #!")
$

Describe the solution you'd like

When datafusion-cli is run on a file it should ignore the first line of the file if it begins with '#!' as the first two characters.

I believe no valid SQL file can begin with '#!'. Therefore it is safe to make this rule. If for some reason someone wanted to begin a statement with #!, they can put a return at the beginning of the file. '#!' must be the first two characters of the file to be a valid hashbang.

Describe alternatives you've considered

No response

Additional context

https://en.wikipedia.org/wiki/Shebang_(Unix)

[PsiACE]

I made a very small modification to address this issue. Additionally, for this scenario, if you are using bash, you can try the following style.

--() { :; }; exec datafusion-cli -f "$0"
SELECT 1;

Ref: https://rosettacode.org/wiki/Multiline_shebang#PostgreSQL