clp-s: Add support for using "?" in wildcard queries on single-word values (resolves #392).

[wraymo]

References

#392

Description

This PR adds the feature requested in #392. Now we can use ? to match an arbitrary single character on a variable field. However, there are still two limitations.

• Queries with a ? in the middle return empty results. This should be fixes by clp-s: Handle cases where clp-string query-parsing falls back to decompress + match (fixes #403). #407.
• Queries combined with ?(*) and \?(\*) cannot work well. This will be fixed later.

Validation performed

Compressed the logs below

{"a": "Hello world?", "b": "wwoorr\lldd", "c":"*world"}
{"a": "Hello *", "b": "?world"}

The queries below all returned correct results

b: \?world
b: ?world
b: ?wo?ld
b: wwoorr\\lldd
c: ?world
c: \*world

[gibber9809]

I know it's wrong in the original code as well, but could we fix this to handle escaped * and ? properly as well? This should really be checking just for unescaped * or ?, and ignoring them when they're escaped. E.g. var\? is just a variable string.

[wraymo]

Right

[gibber9809]

Similar to my other comment we should be handling escaping here as well. Specifically if there are no unescaped * or ? we should enter this code block, process the string to unescape escaped characters as necessary, then do the precise search.

E.g. \\var\? should enter this code block and get unescaped to become \var? before performing exact match.

[wraymo]

I think the function wildcard_match_unsafe_case_sensitive can handle escape characters correctly. The corner case was mentioned above and will be fixed later.

[gibber9809]

Right, but this code block uses m_var_dict->get_entry_matching_value() which doesn't respect escape characters.

[gibber9809]

Maybe for the condition here we can add a method like

bool has_unescaped_wildchards(std::string const& query) {
    bool escaped{false};
    for (size_t i = 0; i < query.size(); ++i) {
        if ('*' == query[i] || '?' == query[i]) {
            return true;
        }
        if ('\' == query[i]) {
            ++i;
        }
    }
    return false;
}

and replace the if condition with false == has_unescaped_wildcards(query)?

[gibber9809]

LGTM. Maybe just add (fixes #392) to the commit message -- besides that title is good as commit message.

[kirkrodrigues]

How about:

clp-s: Add support for using "?" in wildcard queries on single-word values (fixes #392).

Or since #392 is a feature request rather than a bug, we could use "resolves #392".