Pipeline Components and Applications

  1. Home
  2. Docs
  3. Pipeline Components and Applications
  4. Iglu
  5. Igluctl
  6. Igluctl previous versions
  7. Igluctl (0.4.0)

Igluctl (0.4.0)

Iglu provides a CLI application, called igluctl which allows you to perform most common tasks on Iglu registry. So far, it consists of two subcommands: static and lint, whereas former itself contains three subcommands generate,push and s3cp, so overall structure of commands looks like following:

  • lint – validate set of JSON Schemas for syntax and consistency of their properties
  • static – work with static Iglu registry
    • generate – generate DDLs and migrations (only for Redshift now) from set of JSON Schemas
    • push – push set of JSON Schemas from static registry to full-featured (Scala Registry for example) one
    • s3cp – copy JSONPaths or schemas to S3 bucket

Download the latest Igluctl from Bintray:

$ wget http://dl.bintray.com/snowplow/snowplow-generic/igluctl_0.4.0.zip
$ unzip -j igluctl_0.4.0.zip

Now you can use executable:

$ ./igluctl --help

Note that Igluctl supports only JRE/JDK8. Java 7 is not supported and at least s3cp does not work with Java 9.


igluctl lint validates JSON Schemas.

It is designed to be run against schema registries with a folder structure that follows the iglu-example-schema-registry.

You can validate all the schemas in the registry:

$ /path/to/igluctl lint /path/to/schema/registry/schemas

Alternatively you can validate an individual schema e.g.:

$ /path/to/igluctl lint /path/to/schema/registry/schemas/com.example_company/example_event/jsonschema/1-0-0

Examples of errors that are identified:

  • JSON Schema has inconsistent self-describing information and path on filesystem
  • JSON Schema has invalid $schema keyword. It should be always set to iglu-specific, while users tend to set it to Draft v4 or even to self-referencing Iglu URI
  • JSON Schema is invalid against its standard (empty required, string maximum and similar)
  • JSON Schema contains properties which contradict each other, like {"type": "integer", "maxLength": 0} or {"maximum": 0, "minimum": 10'}. These schemas are inherently useless as for some valiators there is no JSON instance they can validate

The above cases can very hard to spot without a specialized tool as they are still valid JSONs and in last case it is even valid JSON Schemas – so will validate against a standard JSON schema validator.

lint has two options:

  • --skip-warnings which will omit warnings like unknown properties. However we strongly advise to not use this option.
  • --skip-checks which will lint without specified linters, given comma separated. To see available linters and their explanations, $ /path/to/igluctl --help

Note: --severityLevel option is deprecated and removed as of version 0.4.0.

Below are two groups of linters; allowed to be skipped and not allowed to be skipped. By default, all of them are enabled but igluctl users can skip any combination of rootObjectunknownFormatsnumericMinMaxstringLengthoptionalNulldescription through --skip-checks.

Igluctl let you skip below checks:

rootObjectCheck that root of schema has object type and contains properties
unknownFormatsCheck that schema doesn’t contain unknown formats
numericMinMaxCheck that schema with numeric type contains both minimum and maximum properties
stringLengthCheck that schema with string type contains maxLength property or other ways to extract max length
optionalNullCheck that non-required fields have null type
descriptionCheck that property contains description

A sample usage could be as following:

$ /path/to/igluctl lint --skip-checks description,rootObject /path/to/schema/registry/schemas

Note that linter names are case sensitive

Igluctl also includes many checks proving that schemas doesn’t have conflicting expectations (such as minimum value bigger than maximum). Schemas with such expectations are valid according to specification, but do not make any sense in real-world use cases. These checks are mandatory and cannot be disabled.

igluctl lint will exit with status code 1 if encounter at least one error.

static generate

igluctl static generate generates corresponding Redshift DDL files (CREATE TABLE statements) and migration scripts (ALTER TABLE statements).

This command previously was a part of Schema Guru and was known as schema-guru ddl, but has been moved into iglu in r5 release.

$ ./igluctl static generate $INPUT

You also can specify directory for output (current dir is used as default):

$ ./igluctl static generate --output $DDL_DIR $INPUT

Generating migration Redshift table scripts to accommodate updated schema versions

If an input directory is specified with several self-describing JSON schemas with a single REVISION, Igluctl will generate migration scripts to update (ALTER) Redshift tables for older schema versions to support the latest schema version.

For example, having the following Self-describing JSON Schemas as an input:

  • schemas/com.acme/click_event/1-0-0
  • schemas/com.acme/click_event/1-0-1
  • schemas/com.acme/click_event/1-0-2

Igluctl will generate the following migration scripts:

  • sql/com.acme/click_event/1-0-0/1-0-1 to alter table from 1-0-0 to 1-0-1
  • sql/com.acme/click_event/1-0-0/1-0-2 to alter table from 1-0-0 to 1-0-2
  • sql/com.acme/click_event/1-0-1/1-0-2 to alter table from 1-0-1 to 1-0-2

This migrations (and all subsequent table definitions) are aware of column order and will ensure that new columns are added at the end of the table definition. This means that the tables can be updated in-place with single ALTER TABLE statements.

NOTE: migrations support is in early beta. Only single alter-table case is supported, particularly “add optional field”.

Handling union types

One of the more problematic scenarios to handle when generating Redshift table definitions is handling UNION field types e.g. ["integer", "string"].

How should these be represented in SQL DDL? It’s a tough question and we believe there is no perfect solution.

Igluctl provides two options:

  1. Union types will be transformed as most general. In the above example (union of an integer and string type) the corresponding Redshift column will be a VARCHAR(4096). This is the default behaviour.
  2. Alternative approach: split the column with product types into separate ones with it’s types as postfix. In the case of a field with type ["string", "integer"] two corresponding columns will be generated in Redshift: model_string and model_integer. To get Igluctl to split columns add the following flag: --split-product-types.

Missing schema versions

static generate command will check versions of schemas inside input as following:

  • If user specified folder and one of schemas has no 1-0-0 or misses any other schemas in between (like it has 1-0-0 and 1-0-2) – refuse to do anything (but proceed with –force option)
  • If user specified full path to file with schema and this file is not 1-0-0 – just print a warning
  • If user specified full path to file with schema and it is 1-0-0 – all good

Other options

If you’re not a Snowplow Platform user, don’t use Self-describing Schema or just don’t want anything Iglu-specific, you can produce raw DDL:

$ ./igluctl static generate --raw $INPUT

But bear in mind that Self-describing Schemas bring many benefits. For example, raw Schemas will not preserve an order for your columns (it’s just impossible as it doesn’t know about previous revisions) and also you will not have a migrations.

You may also want to get JSONPaths file for Redshift’s COPY command. It will place jsonpaths dir alongside with sql:

$ ./igluctl static generate --with-json-paths $INPUT

If there’s no clues about string length (e.g. maxLength specifications in the schema), Igluctl will set the length of VARCHAR columns to 4096 by default . You can also specify this your own VARCHAR size used by default:

$ ./igluctl static generate --varchar-size 32 $INPUT

You can also specify Redshift Schema for your table (don’t confuse database schema and schemas like JSON). For non-raw mode atomic used as default.

$ ./igluctl static generate --raw --dbschema business $INPUT

Some users do not full rely on Igluctl for DDL generation and edit their DDLs manually. By default, Igluctl will not override your files (either DDLs and migrations) if user made any significant changes (comments and whitespaces are not significant). Instead Igluctl will print warning that file has been changed manually. To change this behavior you may specify --force flag.

$ ./igluctl static generate --force $INPUT

It is possible to forget about ownership once table is created. It could be achieved within igluctl as following.

$ ./igluctl static generate $INPUT --set-owner <owner>

igluctl also has an option --no-header which will not place header comments into output DDL.

$ ./igluctl static generate $INPUT --no-header

static push

igluctl static push publishes schemas stored locally to a remote Scala Iglu Registry.

It accept three required arguments:

  • host – Scala Iglu Registry host name or IP address with optional port and endpoint. It should conform pattern host:port/path (or just hostwithout http:// prefix.
  • apikey – master API key, used to create temporary write and read keys
  • path – path to your static registry

Also it accepts optional --public argument which makes schemas available without apikey header.

$ ./igluctl static push /path/to/static/registry iglu.acme.com:80/iglu-server f81d4fae-7dec-11d0-a765-00a0c91e6bf6

static s3cp

igluctl static s3cp is basically a wrapper around AWS Java SDK and closely follows semantics of aws s3 cp command from AWS CLI (you don’t have to install it!). It enables you to upload JSON Schemas or any other files to chosen S3 bucket. Note that it doesn’t perform any syntax/linting checks as it suited to work like simple cp command.

igluctl static s3cp accepts two required arguments and several options:

  • input – path to your files. Required.
  • bucket – S3 bucket name. Required.
  • s3path – optional S3 path to prepend your input root. Usually you don’t need it.
  • accessKeyId – your AWS Access Key Id. This may or or may not be required, depending on your preferred authentication option.
  • secretAccessKey – your AWS Secret Access Key. This may or or may not be required, depending on your preferred authentication option.
  • profile – your AWS profile name. This may or or may not be required, depending on your preferred authentication option.
  • region – AWS S3 region. Default: us-west-2

igluctl static s3cp tries to closely follow AWS CLI authentication process. First it checks if profile name or accessKeyId/secretAccessKey pair provided and uses it. If neither of above provided – it looks into AWS_ACCESS_KEY_IDAWS_SECRET_ACCESS_KEY environment variables. If above aren’t available as well – it ~/.aws/config file. If all above failed – it exits with error.