21 KiB

Raw Blame History

layout	title
doc_page	Transforming Dimension Values

Transforming Dimension Values

The following JSON fields can be used in a query to operate on dimension values.

DimensionSpec

DimensionSpecs define how dimension values get transformed prior to aggregation.

Default DimensionSpec

Returns dimension values as is and optionally renames the dimension.

{
  "type" : "default",
  "dimension" : <dimension>,
  "outputName": <output_name>,
  "outputType": <"STRING"|"LONG"|"FLOAT">
}

When specifying a DimensionSpec on a numeric column, the user should include the type of the column in the outputType field. If left unspecified, the outputType defaults to STRING.

Please refer to the Output Types section for more details.

Extraction DimensionSpec

Returns dimension values transformed using the given extraction function.

{
  "type" : "extraction",
  "dimension" : <dimension>,
  "outputName" :  <output_name>,
  "outputType": <"STRING"|"LONG"|"FLOAT">,
  "extractionFn" : <extraction_function>
}

outputType may also be specified in an ExtractionDimensionSpec to apply type conversion to results before merging. If left unspecified, the outputType defaults to STRING.

Please refer to the Output Types section for more details.

Filtered DimensionSpecs

These are only useful for multi-value dimensions. If you have a row in druid that has a multi-value dimension with values ["v1", "v2", "v3"] and you send a groupBy/topN query grouping by that dimension with query filter for value "v1". In the response you will get 3 rows containing "v1", "v2" and "v3". This behavior might be unintuitive for some use cases.

It happens because "query filter" is internally used on the bitmaps and only used to match the row to be included in the query result processing. With multi-value dimensions, "query filter" behaves like a contains check, which will match the row with dimension value ["v1", "v2", "v3"]. Please see the section on "Multi-value columns" in segment for more details. Then groupBy/topN processing pipeline "explodes" all multi-value dimensions resulting 3 rows for "v1", "v2" and "v3" each.

In addition to "query filter" which efficiently selects the rows to be processed, you can use the filtered dimension spec to filter for specific values within the values of a multi-value dimension. These dimensionSpecs take a delegate DimensionSpec and a filtering criteria. From the "exploded" rows, only rows matching the given filtering criteria are returned in the query result.

The following filtered dimension spec acts as a whitelist or blacklist for values as per the "isWhitelist" attribute value.

{ "type" : "listFiltered", "delegate" : <dimensionSpec>, "values": <array of strings>, "isWhitelist": <optional attribute for true/false, default is true> }

Following filtered dimension spec retains only the values matching regex. Note that listFiltered is faster than this and one should use that for whitelist or blacklist usecase.

{ "type" : "regexFiltered", "delegate" : <dimensionSpec>, "pattern": <java regex pattern> }

Following filtered dimension spec retains only the values starting with the same prefix.

{ "type" : "prefixFiltered", "delegate" : <dimensionSpec>, "prefix": <prefix string> }

For more details and examples, see multi-value dimensions.

Lookup DimensionSpecs

Lookups are an experimental feature.

Lookup DimensionSpecs can be used to define directly a lookup implementation as dimension spec. Generally speaking there is two different kind of lookups implementations. The first kind is passed at the query time like map implementation.

{
  "type":"lookup",
  "dimension":"dimensionName",
  "outputName":"dimensionOutputName",
  "replaceMissingValueWith":"missing_value",
  "retainMissingValue":false,
  "lookup":{"type": "map", "map":{"key":"value"}, "isOneToOne":false}
}

A property of retainMissingValue and replaceMissingValueWith can be specified at query time to hint how to handle missing values. Setting replaceMissingValueWith to "" has the same effect as setting it to null or omitting the property. Setting retainMissingValue to true will use the dimension's original value if it is not found in the lookup. The default values are replaceMissingValueWith = null and retainMissingValue = false which causes missing values to be treated as missing.

It is illegal to set retainMissingValue = true and also specify a replaceMissingValueWith.

A property optimize can be supplied to allow optimization of lookup based extraction filter (by default optimize = true).

The second kind where it is not possible to pass at query time due to their size, will be based on an external lookup table or resource that is already registered via configuration file or/and coordinator.

{
  "type":"lookup",
  "dimension":"dimensionName",
  "outputName":"dimensionOutputName",
  "name":"lookupName"
}

Output Types

The dimension specs provide an option to specify the output type of a column's values. This is necessary as it is possible for a column with given name to have different value types in different segments; results will be converted to the type specified by outputType before merging.

Note that not all use cases for DimensionSpec currently support outputType, the table below shows which use cases support this option:

Query Type	Supported?
GroupBy (v1)	no
GroupBy (v2)	yes
TopN	yes
Search	no
Select	no
Cardinality Aggregator	no

Extraction Functions

Extraction functions define the transformation applied to each dimension value.

Transformations can be applied to both regular (string) dimensions, as well as the special __time dimension, which represents the current time bucket according to the query aggregation granularity.

Note: for functions taking string values (such as regular expressions), __time dimension values will be formatted in ISO-8601 format before getting passed to the extraction function.

Regular Expression Extraction Function

Returns the first matching group for the given regular expression. If there is no match, it returns the dimension value as is.

{
  "type" : "regex",
  "expr" : <regular_expression>,
  "index" : <group to extract, default 1>
  "replaceMissingValue" : true,
  "replaceMissingValueWith" : "foobar"
}

For example, using "expr" : "(\\w\\w\\w).*" will transform 'Monday', 'Tuesday', 'Wednesday' into 'Mon', 'Tue', 'Wed'.

If "index" is set, it will control which group from the match to extract. Index zero extracts the string matching the entire pattern.

If the replaceMissingValue property is true, the extraction function will transform dimension values that do not match the regex pattern to a user-specified String. Default value is false.

The replaceMissingValueWith property sets the String that unmatched dimension values will be replaced with, if replaceMissingValue is true. If replaceMissingValueWith is not specified, unmatched dimension values will be replaced with nulls.

For example, if expr is "(a\w+)" in the example JSON above, a regex that matches words starting with the letter a, the extraction function will convert a dimension value like banana to foobar.

Partial Extraction Function

Returns the dimension value unchanged if the regular expression matches, otherwise returns null.

{ "type" : "partial", "expr" : <regular_expression> }

Search Query Extraction Function

Returns the dimension value unchanged if the given SearchQuerySpec matches, otherwise returns null.

{ "type" : "searchQuery", "query" : <search_query_spec> }

Substring Extraction Function

Returns a substring of the dimension value starting from the supplied index and of the desired length. Both index and length are measured in the number of Unicode code units present in the string as if it were encoded in UTF-16. Note that some Unicode characters may be represented by two code units. This is the same behavior as the Java String class's "substring" method.

If the desired length exceeds the length of the dimension value, the remainder of the string starting at index will be returned. If index is greater than the length of the dimension value, null will be returned.

{ "type" : "substring", "index" : 1, "length" : 4 }

The length may be omitted for substring to return the remainder of the dimension value starting from index, or null if index greater than the length of the dimension value.

{ "type" : "substring", "index" : 3 }

Strlen Extraction Function

Returns the length of dimension values, as measured in the number of Unicode code units present in the string as if it were encoded in UTF-16. Note that some Unicode characters may be represented by two code units. This is the same behavior as the Java String class's "length" method.

null strings are considered as having zero length.

{ "type" : "strlen" }

Time Format Extraction Function

Returns the dimension value formatted according to the given format string, time zone, and locale.

For __time dimension values, this formats the time value bucketed by the aggregation granularity

For a regular dimension, it assumes the string is formatted in ISO-8601 date and time format.

format : date time format for the resulting dimension value, in Joda Time DateTimeFormat, or null to use the default ISO8601 format.
locale : locale (language and country) to use, given as a IETF BCP 47 language tag, e.g. en-US, en-GB, fr-FR, fr-CA, etc.
timeZone : time zone to use in IANA tz database format, e.g. Europe/Berlin (this can possibly be different than the aggregation time-zone)
granularity : granularity to apply before formatting, or omit to not apply any granularity.
asMillis : boolean value, set to true to treat input strings as millis rather than ISO8601 strings. Additionally, if format is null or not specified, output will be in millis rather than ISO8601.

{ "type" : "timeFormat",
  "format" : <output_format> (optional),
  "timeZone" : <time_zone> (optional, default UTC),
  "locale" : <locale> (optional, default current locale),
  "granularity" : <granularity> (optional, default none) },
  "asMillis" : <true or false> (optional) }

For example, the following dimension spec returns the day of the week for Montréal in French:

{
  "type" : "extraction",
  "dimension" : "__time",
  "outputName" :  "dayOfWeek",
  "extractionFn" : {
    "type" : "timeFormat",
    "format" : "EEEE",
    "timeZone" : "America/Montreal",
    "locale" : "fr"
  }
}

Time Parsing Extraction Function

Parses dimension values as timestamps using the given input format, and returns them formatted using the given output format.

Note, if you are working with the __time dimension, you should consider using the time extraction function instead instead, which works on time value directly as opposed to string values.

If "joda" is true, time formats are described in the Joda DateTimeFormat documentation. If "joda" is false (or unspecified) then formats are described in the SimpleDateFormat documentation. In general, we recommend setting "joda" to true since Joda format strings are more common in Druid APIs and since Joda handles certain edge cases (like weeks and weekyears near the start and end of calendar years) in a more ISO8601 compliant way.

If a value cannot be parsed using the provided timeFormat, it will be returned as-is.

{ "type" : "time",
  "timeFormat" : <input_format>,
  "resultFormat" : <output_format>,
  "joda" : <true, false> }

Javascript Extraction Function

Returns the dimension value, as transformed by the given JavaScript function.

For regular dimensions, the input value is passed as a string.

For the __time dimension, the input value is passed as a number representing the number of milliseconds since January 1, 1970 UTC.

Example for a regular dimension

{
  "type" : "javascript",
  "function" : "function(str) { return str.substr(0, 3); }"
}

{
  "type" : "javascript",
  "function" : "function(str) { return str + '!!!'; }",
  "injective" : true
}

A property of injective specifies if the javascript function preserves uniqueness. The default value is false meaning uniqueness is not preserved

Example for the __time dimension:

{
  "type" : "javascript",
  "function" : "function(t) { return 'Second ' + Math.floor((t % 60000) / 1000); }"
}

JavaScript-based functionality is disabled by default. Please refer to the Druid JavaScript programming guide for guidelines about using Druid's JavaScript functionality, including instructions on how to enable it.

Registered lookup extraction function

Lookups are a concept in Druid where dimension values are (optionally) replaced with new values. For more documentation on using lookups, please see Lookups. The "registeredLookup" extraction function lets you refer to a lookup that has been registered in the cluster-wide configuration.

An example:

{
  "type":"registeredLookup",
  "lookup":"some_lookup_name",
  "retainMissingValue":true
}

It is illegal to set retainMissingValue = true and also specify a replaceMissingValueWith.

A property of injective can override the lookup's own sense of whether or not it is injective. If left unspecified, Druid will use the registered cluster-wide lookup configuration.

A property optimize can be supplied to allow optimization of lookup based extraction filter (by default optimize = true). The optimization layer will run on the broker and it will rewrite the extraction filter as clause of selector filters. For instance the following filter

{
    "filter": {
        "type": "selector",
        "dimension": "product",
        "value": "bar_1",
        "extractionFn": {
            "type": "registeredLookup",
            "optimize": true,
            "lookup": "some_lookup_name"
        }
    }
}

will be rewritten as the following simpler query, assuming a lookup that maps "product_1" and "product_3" to the value "bar_1":

{
   "filter":{
      "type":"or",
      "fields":[
         {
            "filter":{
               "type":"selector",
               "dimension":"product",
               "value":"product_1"
            }
         },
         {
            "filter":{
               "type":"selector",
               "dimension":"product",
               "value":"product_3"
            }
         }
      ]
   }
}

A null dimension value can be mapped to a specific value by specifying the empty string as the key in your lookup file. This allows distinguishing between a null dimension and a lookup resulting in a null. For example, specifying {"":"bar","bat":"baz"} with dimension values [null, "foo", "bat"] and replacing missing values with "oof" will yield results of ["bar", "oof", "baz"]. Omitting the empty string key will cause the missing value to take over. For example, specifying {"bat":"baz"} with dimension values [null, "foo", "bat"] and replacing missing values with "oof" will yield results of ["oof", "oof", "baz"].

Inline lookup extraction function

Lookups are a concept in Druid where dimension values are (optionally) replaced with new values. For more documentation on using lookups, please see Lookups. The "lookup" extraction function lets you specify an inline lookup map without registering one in the cluster-wide configuration.

Examples:

{
  "type":"lookup",
  "lookup":{
    "type":"map",
    "map":{"foo":"bar", "baz":"bat"}
  },
  "retainMissingValue":true,
  "injective":true
}

{
  "type":"lookup",
  "lookup":{
    "type":"map",
    "map":{"foo":"bar", "baz":"bat"}
  },
  "retainMissingValue":false,
  "injective":false,
  "replaceMissingValueWith":"MISSING"
}

The inline lookup should be of type map.

The properties retainMissingValue, replaceMissingValueWith, injective, and optimize behave similarly to the registered lookup extraction function.

Cascade Extraction Function

Provides chained execution of extraction functions.

A property of extractionFns contains an array of any extraction functions, which is executed in the array index order.

Example for chaining regular expression extraction function, javascript extraction function, and substring extraction function is as followings.

{
  "type" : "cascade", 
  "extractionFns": [
    { 
      "type" : "regex", 
      "expr" : "/([^/]+)/", 
      "replaceMissingValue": false,
      "replaceMissingValueWith": null
    },
    { 
      "type" : "javascript", 
      "function" : "function(str) { return \"the \".concat(str) }" 
    },
    { 
      "type" : "substring", 
      "index" : 0, "length" : 7 
    }
  ]
}

It will transform dimension values with specified extraction functions in the order named. For example, '/druid/prod/historical' is transformed to 'the dru' as regular expression extraction function first transforms it to 'druid' and then, javascript extraction function transforms it to 'the druid', and lastly, substring extraction function transforms it to 'the dru'.

String Format Extraction Function

Returns the dimension value formatted according to the given format string.

{ "type" : "stringFormat", "format" : <sprintf_expression>, "nullHandling" : <optional attribute for handling null value> }

For example if you want to concat "[" and "]" before and after the actual dimension value, you need to specify "[%s]" as format string. "nullHandling" can be one of nullString, emptyString or returnNull. With "[%s]" format, each configuration will result [null], [], null. Default is nullString.

Upper and Lower extraction functions.

Returns the dimension values as all upper case or lower case. Optionally user can specify the language to use in order to perform upper or lower transformation

{
  "type" : "upper",
  "locale":"fr"
}

or without setting "locale" (in this case, the current value of the default locale for this instance of the Java Virtual Machine.)

{
  "type" : "lower"
}

Bucket Extraction Function

Bucket extraction function is used to bucket numerical values in each range of the given size by converting them to the same base value. Non numeric values are converted to null.

size : the size of the buckets (optional, default 1)
offset : the offset for the buckets (optional, default 0)

The following extraction function creates buckets of 5 starting from 2. In this case, values in the range of [2, 7) will be converted to 2, values in [7, 12) will be converted to 7, etc.

{
  "type" : "bucket",
  "size" : 5,
  "offset" : 2
}

21 KiB Raw Blame History