jubaconv

Synopsis

jubaconv [options ...]

Description

jubaconv is a tool to test your fv_converter configuration.

jubaconv simulates the internal behavior of fv_converter and displays the result of conversion on the command-line.

As jubaconv only processes 1 record, global_weight of string feature extraction rules (string_rules) will not be calculated. Use Weight if you need a simulation including global_weight.

Options

• [] indicates the default value.

-i <format>, --input-format <format>

Format of the input. [json]

<format> must be one of datum or json.

-o <format>, --output-format <format>

Format of the output. [fv]

<format> must be one of fv, datum or json.

-c <config>, --conf <config>

Jubatus server configuration file in JSON.

This option must be given only if fv is specified for -o.

• Note: If the input format is json, we have to add a slash symbol / at the start of key in jubatus server config files.

  • If the input format is datum: "num_rules" : [{"key": "number", "type": "num"}]
  • If the input format is json: "num_rules" : [{"key": "/number", "type": "num"}]

File Formats

Input

Supported formats are datum and json. See the Examples section for details.

• datum is a Datum-styled JSON data structure. The root must have 3 keys (string_values, num_values and binary_values), and whose values must be a flat object, whose keys are string and values are string (for string_values and binary_values) or integer/float (for num_values).
• json is an arbitrary JSON data structure. JSON data structure will be flatten to key-value format by joining keys with / separator.

Output

Supported formats are fv, datum and json.

• fv prints the feature vectors extracted from the input record by fv_converter.
• datum prints the input record converted to Datum as Datum-styled JSON data structure.
• json pretty-prints the input record without any conversion. This can only be used if the input format is json.

Examples

$ cat data.json
{ "message": "hello world", "age": 31 }

$ jubaconv -i json -o fv -c /opt/jubatus/share/jubatus/example/config/classifier/pa.json < data.json
/message$hello world@str#bin/bin: 1
/age@num: 31

$ cat datum.json
{
  "string_values": {
    "hello": "world"
  },
  "num_values": {
    "age": 31
  },
  "binary_values": {
  }
}

$ jubaconv -i datum -o fv -c /opt/jubatus/share/jubatus/example/config/classifier/pa.json < datum.json
hello$world@str#bin/bin: 1
age@num: 31