Rule-based JSON/YAML validator using JSON schemas
salty-dog is a tool to validate JSON and YAML data using JSON schema rules. It can filter elements and validate
select parts of the document, supports multiple documents in the same stream or file, and can insert defaults during
validation.
salty-dog is distributed as both a Docker container and an npm package, so it can be installed or pulled:
# docker image
> docker pull ssube/salty-dog:master
# npm project install
> yarn add -D salty-dog
# npm global install
> yarn global add salty-dog
Note: while the container is the preferred way of running salty-dog, it has a serious limitation: docker run
combines stdout and stderr, making it difficult to separate logs and the output document. Writing either the logs
or data to a file works around this.
To download, validate, and apply a Kubernetes resource:
> curl https://raw.githubusercontent.com/ssube/k8s-shards/master/roles/apps/gitlab/server/templates/ingress.yml | \
salty-dog \
--rules rules/kubernetes.yml \
--source - \
--tag kubernetes | \
kubectl apply --dry-run -f -
...
{"name":"salty-dog","hostname":"cerberus","pid":7860,"level":30,"msg":"all rules passed","time":"2019-06-16T02:04:37.797Z","v":0}
ingress.extensions/gitlab created (dry run)
salty-dog is written in Typescript and requires make, node, and yarn to build.
> git clone git@github.com:ssube/salty-dog.git
> cd salty-dog
> make
After building, run with node out/index.js or install globally with make yarn-global.
make targets are provided for some example arguments:
> curl https://raw.githubusercontent.com/ssube/k8s-shards/master/roles/apps/gitlab/server/templates/ingress.yml | \
make run-stream \
1> >(kubectl apply --dry-run -f -) \
2> >($(yarn bin)/bunyan)
...
[2019-06-16T03:23:56.645Z] INFO: salty-dog/8015 on cerberus: all rules passed
ingress.extensions/gitlab created (dry run)
This method does not require the usual dependencies to be installed, only docker itself.
Build with Docker:
# Stretch
docker run --rm -v "$(pwd):/salty-dog" -w /salty-dog node:16-stretch make
docker build -t salty-dog:stretch -f Dockerfile.stretch .
# Alpine
docker run --rm -v "$(pwd):/salty-dog" -w /salty-dog node:16-alpine sh -c "apk add build-base && make"
docker build -t salty-dog:alpine -f Dockerfile.alpine .
salty-dog is distributed as a docker container and an npm package.
While the container is the preferred way of running salty-dog, it has one limitation: docker run combines
stdout and stderr, making it impossible to separate logs and the output document. Writing either the logs or dest
to a file works around this.
To run the Docker container: docker run --rm ssube/salty-dog:master
The latest semi-stable image is published to ssube/salty-dog:master. Containers are published based on both Alpine
Linux and Debian (currently Stretch). All of the available tags are listed here.
Rules are provided in the image at /salty-dog/rules. To use custom rules in the container, mount them with
-v $(pwd)/rules:/salty-dog/rules:ro and load them with --rules /rules/foo.yml.
The ssube/salty-dog container image can be run normally or interactively.
To validate a file or input normally:
> docker run --rm ssube/salty-dog:master --help
You can also launch a shell within the container, using local rules:
> docker run \
--rm \
-it \
--entrypoint bash \
ssube/salty-dog:master
salty-dog is also published as an npm package with a binary, so it can
be used as a CLI command or programmatically.
To install salty-dog for the current project:
> yarn add -D salty-dog
> $(yarn bin)/salty-dog --help
It is also possible to install salty-dog globally, rather than within a project. However, this is
not recommended.
> yarn global add salty-dog
> export PATH="${PATH}:$(yarn global bin)"
> salty-dog --help
TODO
Rules are the core of salty-dog validation. Each rule has a JSON schema used to check the data, an optional
filter to skip some data, and a name and description.
For example:
name: salty-dog-gitlab-ci
rules:
- name: gitlab-stages
desc: should specify stages
level: info
tags:
- gitlab
- optional
check:
type: object
required: [stages]
properties:
stages:
type: array
items:
type: string
The complete rule format is documented here.
Rules combine a jsonpath expression and JSON schema to select and validate the document.
The rule’s select expression is used to select nodes that should be validated, which are filtered, then checked.
The structure of rule files and the rules within them are documented here.
Rules can be loaded from a file, module, or path.
To load a file by name, --rule-file foo.yml. This will accept any extension.
To load a module, --rule-module foo. The required module exports are documented here.
To load a path, --rule-path foo/. This will recursively load any files matching *.+(json|yaml|yml).
All rules are disabled by default and must be enabled by name, level, or tag.
To enable a single rule by name, --include-name foo-rule.
To enable a group of rules by level, --include-level warn.
To enable a group of rules by tag, --include-tag foo.
To validate the rules in the rules/ directory using the meta-rules:
> make test-rules
...
{"name":"salty-dog","hostname":"cerberus","pid":29403,"level":30,"msg":"all rules passed","time":"2019-06-16T00:56:55.132Z","v":0}
salty-dog outputs two streams: valid input data and error logs.
Valid input data is written back out to stdout, allowing salty-dog to be used inline with piped shell commands.
For example, to validate a kubernetes resource before applying it:
> curl https://raw.githubusercontent.com/ssube/k8s-shards/master/roles/apps/gitlab/server/templates/ingress.yml | \
salty-dog \
--rules rules/kubernetes.yml \
--source - \
--tag kubernetes | \
kubectl apply --dry-run -f -
...
{"name":"salty-dog","hostname":"cerberus","pid":7860,"level":30,"msg":"all rules passed","time":"2019-06-16T02:04:37.797Z","v":0}
ingress.extensions/gitlab created (dry run)
Properties that appear in the schema with the default key set will be added to each element as it is checked. Rules
apply in order, as do their defaults.
salty-dog uses node-bunyan for logging and prints structured JSON output.
Logs can be pretty-printed by redirecting stderr through bunyan itself or jq, both of which are installed in
the salty-dog container.
To filter out error messages, then format the errors they contain:
> salty-dog | jq 'select(.level > 30) | {msg: .msg, errors: try (.errors[] | .msg), rule: try (.rule.desc)}'
[
"rule failed",
"containers must have complete resources specified"
]
[
"some rules failed",
"/resources/limits must have required property 'cpu' at item 0 of $.spec.template.spec.containers[*] for kubernetes-resources",
null
]