CPP-747 fixing inconsistency in USE statements

[derek9gag]

The USE <keyspace> statements are inconsistently quoted in different places. It is impossible to use a quoted keyspace (for example keyspace with upper-case characters, see CPP-747).

This PR fixes a regression from v2.9.0. Keyspaces with uppercases likely never worked since v2.10.0 (9292b53).

This also resolves PHP-224.

See these lines:

• https://github.com/datastax/cpp-driver/blob/bbbbd7bc3eaba1b10ad8ac6f53c41fa93ee718db/src/prepare_host_handler.cpp#L183
• https://github.com/datastax/cpp-driver/blob/bbbbd7bc3eaba1b10ad8ac6f53c41fa93ee718db/src/pooled_connection.cpp#L40
• https://github.com/datastax/cpp-driver/blob/bbbbd7bc3eaba1b10ad8ac6f53c41fa93ee718db/src/connector.cpp#L265

[mpenick]

I think this will break using case intensive keyspaces from the C API.

There are two cases:

1. User provided keyspaces from the C API these take the form caseinsensitive or \"CaseSensitive\"
2. Cassandra provided keyspaces (from its metadata) which are will be sent as having the correct case, but i think won't be wrap in the double quotes when case senstivie. I need to verify this.

So it's possible that we should use the form "USE " + keyspace, but properly wrap Cassandra provided keyspaces in double quotes when needed?

[mpenick]

The other possibility is that we just need to use "USE \"" + keyspace + "\"" in the right spots e.g. PrepareHostHandler::SetKeyspaceCallback, but not the other two spots as long as we can prove that its keyspace values never come from the C API. We'd also have to prove that the other two spot don't take values from Cassandra provided keyspaces.

[derek9gag]

I have tried those using cqlsh, assuming we have a "CaseSensitive" and a case_insensitive keyspaces:

• USE "CaseSensitive"; works
• USE CaseSensitive; fails
• USE "case_insensitive"; works
• USE case_insensitive; works
• USE CASE_INSENSITIVE; works
• USE "CASE_INSENSITIVE"; fails

Cassandra provided keyspaces such as system and system_auth are the same as an user case_insensitive keyspace.

Always quoting the namespace only fails when we specify the wrong casing (e.g. USE "System";).

When we retrieve the metadata, the keyspaces will be have the correct casing like system and system_auth, and adding quotes around will work.

[derek9gag]

I see there is one case-sensitive Cassandra-provided table (system."IndexInfo"), but I have not seen and am not aware of any case-sensitive Cassandra-provided keyspaces.

[derek9gag]

I tried listing the keyspace names from CassSession schema. The keyspace names are not quoted. I did this via the PHP wrapper:

Listing keyspace names in $session->schema()->keyspaces():
 UpperCase
 lowercase
 system
 system_auth
 system_distributed
 system_schema
 system_traces

I understand that we can do it the other way round (i.e. no quoting in USE <keyspace>;), but I would say that is not the correct approach. First it is counter-intuitive to use, as the quote characters are not part of the keyspace name. Secondly, if you have a program that loops through each keyspace in schema, it will fail on the CaseSensitive ones.

[mpenick]

The keyspace can't be double quoted in connector.cpp (https://github.com/datastax/cpp-driver/pull/466/files#diff-0fc019678d7d3eecf2cb341a50df9945R265) because it breaks a few cases when you connect with a keyspace.

That's what I was attempting to explain above.

  { // Works
    CassError rc;
    CassSession* session = cass_session_new();
    CassFuture* future = cass_session_connect_keyspace(session, cluster,  "case_insensitive");
    rc = cass_future_error_code(future);
    assert(rc == CASS_OK && "Same case for insensitive should work");
    cass_future_free(future);
    cass_session_free(session);
  }

  { // Should work, but doesn't work: Received error response 'Keyspace 'Case_Insensitive' does not exist' (0x02002200)
    CassError rc;
    CassSession* session = cass_session_new();
    CassFuture* future = cass_session_connect_keyspace(session, cluster,  "Case_Insensitive");
    rc = cass_future_error_code(future);
    assert(rc == CASS_OK && "Different case for insensitive should work");
    cass_future_free(future);
    cass_session_free(session);
  }

  { // Fails (because it's double, double quoted): 'line 1:4 no viable alternative at input '""' (USE [""]...)' (0x02002000)
    CassError rc;
    CassSession* session = cass_session_new();
    CassFuture* future = cass_session_connect_keyspace(session, cluster, "\"CaseSensitive\"");
    rc = cass_future_error_code(future);
    assert(rc == CASS_OK && "Same case with quotes for sensitive should work");
    cass_future_free(future);
    cass_session_free(session);
  }

  { // Works, but shouldn't (it should require double quotes to work)
    CassError rc;
    CassSession* session = cass_session_new();
    CassFuture* future = cass_session_connect_keyspace(session, cluster, "CaseSensitive");
    rc = cass_future_error_code(future);
    assert(rc != CASS_OK && "Same case without quotes for sensitive should not work");
    cass_future_free(future);
    cass_session_free(session);
  }

[mpenick]

I have a potential fix on a branch here: 72ef57d

[derek9gag]

I see. I checked out your branch and it works fine for my use-case.

[mpenick]

Added a follow up PR (w/ additional tests) #467.

[derek9gag]

Thank you!

[derek9gag]

See #467