Incorrect backslash treatment in string literals in DataFusion CLI

[findepi]

In standard SQL, the \ character has no special meaning in '...' varchar literals.

In PostgreSQL

postgres=# SELECT '\', '\\'
postgres-# ;
 ?column? | ?column?
----------+----------
 \        | \\
(1 row)

in DataFusion sqllogictest the behavior is the same:

query T
SELECT '\'
----
\

query T
SELECT '\\'
----
\\

query T
SELECT '\\\'
----
\\\

query T
SELECT '\\\\'
----
\\\\

However, DataFusion CLI behaves differently

$ cargo run
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.15s
     Running `target/debug/datafusion-cli`
DataFusion CLI v43.0.0
> SELECT '\\';
+-----------+
| Utf8("\") |
+-----------+
| \         |
+-----------+
1 row(s) fetched.
Elapsed 0.049 seconds.

> SELECT '\';  🤔 Invalid statement: SQL error: TokenizerError("unsupported escape char: '\\''")

Given that DataFusion CLI is used a lot to test and verify DataFusion's behavior, it's super important for the CLI to behave correctly with respect to its input.

[alamb]

BTW @pmcgleenon hit this when updating ClickBench as well:

ClickHouse/ClickBench#301

[Lordworms]

take

[alamb]

@Lordworms has a proposed PR here to fix this:

• Do not unescape backslashes in datafusion-cli #14844

However it seems like the way to get consistent behavior with sqllogictest (and psql) is to avoid unescaping in datafusion-cli 🤔

Any thoughts on this?

[alamb]

@Lordworms has a proposed PR here to fix this:

• Do not unescape backslashes in datafusion-cli #14844

However it seems like the way to get consistent behavior with sqllogictest (and psql) is to avoid unescaping in datafusion-cli 🤔

Any thoughts on this?

Documentatiion in

• Document SQL literal syntax and escaping #14934