Page tree

Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Published by Scroll Versions from space DEV and version r0682

D toc

As needed, you can deploy custom data types into the

D s platform
rtrue
, in which type validation is performed against regular expressions that you specify. This method is most useful for validating against patterns, as opposed to specific values.

Warning

After a custom type has been added, it cannot be removed or disabled. Please verify your regular expression before saving the type.

Custom Types Location

On the server hosting the

D s platform
, type definitions are stored in the following directory:

Code Block
/opt/trifacta/node_modules/jsdata/type-packs/trifacta

This directory is referenced as $CUSTOM_TYPE_DIR in the steps below.

Warning

Before you begin creating custom data types, you should backup the type-packs/trifacta directory to a location outside of your

D s item
itemdeployment
.


Info

NOTE: The trifacta-extras directory in the type-packs directory contains experimental custom data types. These data types are not officially supported. Please use with caution.

Directory contents:


  • The dictionaries sub-directory contains user-defined dictionaries.

    Info

    NOTE: Please use the user interface to interact with your dictionaries. See Custom Type Dialog.

  • The types sub-directory contains individual custom data type definitions, each in a separate file. 

  • The manifest.json file contains a JSON manifest of all of standard and custom types in the system.

Examples

Each custom data type is created and stored in a separate file. The following example file contains a regular expression method for validating data against the set of days of the week:

Code Block
{
  "name": "DayOfWeek",
  "prettyName": "Day of Week",
  "category" : "Date/Time",
  "defaultProbability": 1E-15,
  "testCase": {
    "stripWhitespace": true,
    "regexes": [
      "^(monday|tuesday|wednesday|thursday|friday|saturday|sunday)$",
      "^(mon|tue|wed|thu|fri|sat|sun)$"
    ],
    "probability": 0.001
  }
} 

Parameters:

Parameter NameDescription
name

Internal identifier for the custom type. Must be unique across all standard types and custom types.

Info

NOTE: You should verify that your data type's name value does not conflict with other custom data type names.

prettyNameDisplay name for the custom type.
category

The category to assign to the type. The current categories are displayed within the data type drop-down for each column.

defaultProbabilityAssign a default probability for the custom type. See below.
testCaseThis block contains the regular expression specification to be applied to the column values.
stripWhitespaceWhen set to true, whitespace is removed from any value prior for purposes of validation. The original value is untouched.
regexes

This array contains a set of regular expressions that are used to validate the column values. For a regex type, the column value must match with at least one value among the set of expressions.

Info

NOTE: All match types must be double-escaped in the regex expression. For example, to replicate the \d pattern, you must enter: \\d.

D s product
 implements a version of regular expressions based off of RE2 and PCRE regular expressions.

probability(optional) Assign an incremental change to the probability when a match is found between a value and one of the regular expressions. See Defining probabilities below.
Tip

Tip: In the types sub-directory, you can review the regex-based types that are provided with the

D s platform
. While you should not edit these files directly, they may provide some guidance and some regex tips on how to configure your own custom data types. 

Defining probabilities

For your custom type, the probability values are used to determine the likelihood that matching values indicate that the entire column is of the custom data type.

  • The defaultProbability value specifies the baseline probability that a match between a value and one of the regular expressions indicates that the column is the specified type. On a logarithmic scale, values are typically 1E-15 to 1E-20.
  • When a value is matched to one of the regular expressions, the probability value is used to increment the baseline probability that the next matching value is of the specified type. This value should also be expressed on a logarithmic scale (e.g. 0.001).
  • In this manner, a higher number of matching values increases the probability that the type is also a match to the custom type.

Probabilities become important primarily if you are creating a custom type that is a subset of an existing type. For example, the Email Address custom type is a subset of String type. So, matches for the patterns expressed in the Email Address definition should register a higher probability value than the same incremental for the String type definition.

Tip

Tip: For custom types that are subsets of other, non-String types, you should lower the defaultProbability of the baseline type by a factor of 10 (e.g. 1E-15 to 1E-16) and raise the same probability in the custom type by a factor of 10 (e.g. 1E-14). In this manner, you can give higher probability of matching to these subset types.

Add custom types to manifest

To the $CUSTOM_TYPE_DIR/manifest.json file, you must add the filenames of any custom types that you have created and stored in the types directory:

Code Block
{
  "types": ["bodies-of-water.json", "dayofweek.json"],
  "dictionaries": ["oceans", "seas"]
} 

Enable custom types

To enable use of your custom data types in the

D s platform
, locate and edit enabledSemanticTypes property.

D s config

Info

NOTE: Add your entries to the items that are already present in enabledSemanticTypes. Do not delete and replace entries.

Info

NOTE: Do not use this parameter to attempt to remove specific data types. Removal of the default types is not supported.

Code Block
"webapp.enabledSemanticTypes": [
    "<CustomTypeName1>",
    "<CustomTypeName2>",
    "<CustomTypeNameN>"
]

where:

  • <CustomTypeName1> corresponds to the internal name value for your custom data type.

Register your custom types

To add your custom types to the

D s platform
, run the following command from the js-data directory: 

Code Block
node bin/load-types --manifest ${PATH_TO_MANIFEST_FILE} 

Restart platform

Restart services. See Start and Stop the Platform.

Check for the availability of your types in the column drop-down. See Column Menus.