GetStateByPartialCompositeKey方法异常：非首个复合键查询无返回求助

Ah, I’ve run into this exact issue before—Fabric’s composite key partial matching is strict about prefixes, which trips up a lot of folks. Let’s break down why your query with key3 isn’t working, then walk through practical solutions.

Why Your Current Query Fails #

GetStateByPartialCompositeKey works by matching prefixes of the composite key’s component sequence. Your index is structured as key1~key2~key3~key4~key5, which means valid partial queries must start from the first component and follow sequentially:

• Empty (matches all entries in the index)
• key1 (matches all entries starting with key1)
• key1 + key2 (matches all entries starting with key1~key2)
• And so on up to the full 5 components

When you pass just key3, Fabric looks for composite keys that start with key3—but all your entries start with key1, so no results come back. That’s why your first query with key1 works perfectly.

Solutions to Query by key3 #

1. Create a Dedicated Index for key3 #

The cleanest and most efficient fix is to design a new composite index where key3 is the first component. This lets you use GetStateByPartialCompositeKey directly for key3 queries.

Example: Writing the Index

When you save your main data, also write an entry to the new index. You can store either the main data key (to look up full data later) or the data itself:

// Main data composite key
mainKey, err := stub.CreateCompositeKey("main_index", []string{key1, key2, key3, key4, key5})
if err != nil {
    return shim.Error(err.Error())
}
err = stub.PutState(mainKey, yourDataBytes)
if err != nil {
    return shim.Error(err.Error())
}

// New index for key3 queries
key3IndexKey, err := stub.CreateCompositeKey("index_by_key3", []string{key3, key1, key2, key4, key5})
if err != nil {
    return shim.Error(err.Error())
}
// Store the main key to retrieve full data later
err = stub.PutState(key3IndexKey, []byte(mainKey))
if err != nil {
    return shim.Error(err.Error())
}

Example: Querying the New Index

Now you can query by key3 successfully:

resultsIterator, err := stub.GetStateByPartialCompositeKey("index_by_key3", []string{key3})
if err != nil {
    return shim.Error(err.Error())
}
defer resultsIterator.Close()

var output []map[string]interface{}
for resultsIterator.HasNext() {
    res, err := resultsIterator.Next()
    if err != nil {
        return shim.Error(err.Error())
    }
    // Fetch full data using the stored main key
    mainData, err := stub.GetState(string(res.Value))
    if err != nil {
        return shim.Error(err.Error())
    }
    // Parse and add to output
    var data map[string]interface{}
    json.Unmarshal(mainData, &data)
    output = append(output, data)
}

return shim.Success(json.Marshal(output))

2. Use Range Queries (Advanced) #

If you can’t create a new index for some reason, you can construct a range query that targets the key3 component. Fabric uses \x00 (null byte) to separate composite key components, so you can build start/end keys to match entries where key3 is your target.

For example, your index keys look like:
index\x00key1\x00key2\x00key3\x00key4\x00key5

You’d construct a range from:
index\x00\x00\x00your_target_key3
to:
index\x00\xff\xffyour_target_key3\xff\xff

Then use GetStateByRange to fetch these entries. This is error-prone though—easy to mess up the separator logic—and less efficient than a dedicated index.

3. Use CouchDB Rich Queries (If Available) #

If your peer uses CouchDB instead of LevelDB as the state database, you can skip composite keys entirely for this query. Just store your data as JSON, define a schema that indexes the key3 field, and use GetQueryResult to run a SQL-like query:

query := fmt.Sprintf(`{"selector": {"key3": "%s"}}`, key3)
resultsIterator, err := stub.GetQueryResult(query)
// ... process results as before

Key Notes #

• Maintain Indexes: When updating or deleting data, make sure to update/delete the corresponding index entries to avoid stale data.
• Index Overhead: Each index adds write latency and storage usage—only create indexes for queries you actually need.

内容的提问来源于stack exchange，提问作者Help_needed