Mario Mario - 1 month ago 6
reST (reStructuredText) Question

HBase REST Filter ( SingleColumnValueFilter )

I cannot figure out how to use filters in the HBase REST interface (HBase 0.90.4-cdh3u3). The documentation just gives me a schema definition for a "string", but doesn't show how to use it.

So, I'm able to do this:

curl -v -H 'Content-Type: text/xml' -d '<Scanner startRow="ddo" stopRow="ddp" batch="1024"/>' 'http://hbasegw:8080/table/scanner'


and then retrieve with

curl -s -H "Content-Type: text/xml" http://hbasegw:8080/table/scanner/13293426893883128482b | tidy -i -q -xml


But now I want to use a SingleColumnValueFilter and have to encode that somehow in the XML.
Does anyone have an example for this?

Thanks,
Mario

Answer

Filter fields in the Scanner XML are strings formatted as JSON. Since the JSON for the filter has many quotes in it, I recommend using a separate file for curl's -d parameter, to avoid the single quote.

curl -v -H "Content-Type:text/xml" -d @args.txt http://hbasegw:8080/table/scanner

Where the file args.txt is:

<Scanner startRow="cm93MDE=" endRow="cm93MDg=" batch="1024">
    <filter>
    {
        "latestVersion":true, "ifMissing":true, 
        "qualifier":"Y29sMQ==", "family":"ZmFtaWx5", 
        "op":"EQUAL", "type":"SingleColumnValueFilter", 
        "comparator":{"value":"MQ==","type":"BinaryComparator"}
    }
    </filter>
</Scanner>

How do you discover how the JSON filter string should look like? Here's an easy way through Java code that spits out the stringified filter given a standard Filter object from HBase's Java API.

SingleColumnValueFilter filter = new SingleColumnValueFilter(
    Bytes.toBytes("family"),
    Bytes.toBytes("col1"),
    CompareFilter.CompareOp.EQUAL,
    Bytes.toBytes("1")
);
System.out.println(ScannerModel.stringifyFilter(filter));

Notice that the JSON and the XML require data encoded in Base64. I've tested the above curl command on a table and it worked just fine.

In case you are wondering, yes, the REST API for scanners is not yet as developer-friendly as it can get.