Alphanumeric partial matching fails (e.g., "610" doesn't find "CL610-ABC")

[davlet61]

Hi,

I am experiencing and issue where searching for partial alphanumeric codes results in 0 hits, or empty array.

Consider following words CL610-ABC, Well-7162-XYZ:

• Search "610" → Should find "CL610-ABC" ✅
• Search "7162" → Should find "Well-7162-XYZ" ✅
• Search "10-a" → Should find "CL610-ABC" ✅

Currently none of these have hits. The behavior is a bit strange, if I the search term starts with the target word beginning it works. Or if the mid-word term is only letters, we get some hits (E.g: AB will return CL610-ABC)

I am wondering whether or not I am doing something wrong, or maybe there is a built-in way to enable substring/partial matching for alphanumeric identifiers?

    const orama = create({
        schema: {
            id: "string",
            name: "string",
            type: "enum",
            region_id: "number",
            location: "geopoint",
            sample_names: "string[]",
            well_id: "number",
            well_name: "string",
            fluid_type: "string",
        } as const,
    });

// ...
    const results = await search(orama, {
        term: query,
        limit: 12,
        properties: ["name", "sample_names", "well_name"],
    });

[micheleriva]

Hi @davlet61, this is expected as 610 is a substring of CL610-ABC. The tokenizer splits CL610-ABC into ["CL610", "ABC"] and can't find a string starting with 610.

Doing substring match is feasable, but will drastically increase index size (and therefore memory utilization) and it may not be doable on a browser.

I hope this clarifies things!