Skip to content

Latest commit

 

History

History
294 lines (248 loc) · 11.8 KB

CHANGELOG.md

File metadata and controls

294 lines (248 loc) · 11.8 KB

Changelog for Snow

This document details changes between each release.

See: Keep a Changelog

[0.16.0]

Changed

  • Disabled HTML escaping when printing JSON, in JSON.print. It turned out that escaping was the default.
  • Changed "keywordLocation" and "absoluteKeywordLocation" in the output to "schemaLocation" and "absSchemaLocation", respectively.
  • Updated dependency versions:
    1. ClassGraph: 4.8.90 -> 4.8.102
    2. ICU4J: 68.1 -> 68.2

Fixed

  • Fixed $recursiveRef root lookup. Added additional search for stripped empty fragment in known IDs.
  • Fixed ContentEncoding and ContentMediaType to only assert for Draft-07. It was previously behaving as a validation assertion for Draft 2019-09.

[0.15.0]

Added

  • Added an "Internal APIs" section to the README. This mentions useful classes and methods for use in one's own projects.
  • The test suite runner can now optionally dump all the errors and annotations.
  • Added a new Locator class that encapsulates instance and schema locations. Annotation and Error use this.
  • New JSON.print methods for outputting a JSON element. There is one for an OutputStream and one for a Writer. Coverage, Linter, Main, and Test were changed to use this function for their output.
  • New linter rules:
    1. Added a check for case-insensitive matching for unknown keywords.
    2. Added implied type checks for "default" and "const".
    3. Check for non-unique enums.
    4. Checks for empty "enum", "allOf", "anyOf", and "oneOf".
  • New Results utility class for post-processing annotations and errors into useful results.
  • A rudimentary way to translate ECMA-262 regexes into Java-compatible regexes so that we can pass the optional ECMA-262 regex tests. This is done internally.
  • Added module information:
    • Exporting:
      1. com.qindesign.json.schema
      2. com.qindesign.json.schema.net
    • Requiring transitively:
      1. com.google.gson

Changed

  • Updated to the latest Draft 2019-09 and Draft-06 schemas.
  • Changed the schema resource hierarchy so that it matches the JSON Schema upstream hierarchy.
  • Restructured annotations and errors. Annotation was split into the parameterized Annotation and Error classes. Error replaces the ValidationResult Annotation value.
  • Changed the linter to output JSON.
  • Changed the library used to find Keyword implementation classes. Now using ClassGraph instead of Guava. This somewhat reduces the dependency size.
  • When examining URLs during auto-resolution, the empty path case is now ignored. Before, just a check for an ending "/" was performed. The goal is not to resolve a non-schema file.
  • README and Javadoc updates.
  • Improved auto-resolution:
    1. If checking a URI against the base URI yields nothing, an attempt will now be made to try the URI itself as a URL.
    2. $recursiveRef was added to reference scanning candidates.
    3. Absolute referenced URI's are now added to the list of known URLs in the scan phase; before, any URI's having a scheme or authority were not added to the list of guessable URLs.
  • Renamed package com.qindesign.net to com.qindesign.json.schema.net for more specificity.
  • Updated dependency versions:
    1. ICU4J: 67.1 -> 68.1
  • Updated the Javadocs with proper external links.

Fixed

  • The validator context now knows when it's validating a meta-schema. CoreVocabulary was fixed to ignore the keyword if the current schema isn't being treated as a meta-schema.
  • A check for validator creations throwing an IllegalArgumentException was added to the test suite runner.
  • Because the validator now may add absolute URIs to the list of a schema's known IDs, the context was fixed to first check an ID's base URI, in addition to checking that it has an empty path, before deciding it's the root ID.

[0.14.0]

Added

  • First Maven Central Repository release.
  • Added a JSON.getStringMember convenience method that can get the value of a string-valued object member. It can easily access the title, for example.

Changed

  • Changed the URL and Pattern caches to allow for a non-restricted size. They both remain associated with a specific ValidatorContext instance and are not shared across instances.
  • Improved Gson's misleading syntax error messages. They refer to setting lenient mode to true, even if that's not what is wanted and if setting it to true wouldn't help anyway.
  • Changed the $ref and $recursiveRef error message for target schema failures to the resolved URI of the target schema.
  • Added sorting to both basic and annotation output in Main. The output is sorted by instance location and then by schema location.

Fixed

  • Fixed error pruning for passing schemas to restrict by schema location.
  • Added a missing CONTENT option section to the README.
  • Fixed CoreRecursiveRef to use the correct absolute location when applying the schema.

[0.13.0]

Added

  • Added a Validator constructor for storing information across schema validation for the same schema.
  • Added URI.encodeFragment for easy conversion from JSON Pointers to URI fragments.
  • Added JSONPath.toURIFragmentID to return the URI fragment identifier form.

Changed

  • Annotation and error collection is now specified by passing null or non-null maps instead of the options.
  • Empty annotation elements are now removed before returning from ValidatorContext.apply.
  • Changed all "https://tools.ietf.org/html/rfcXXXX" URLs to "https://www.rfc-editor.org/rfc/rfcXXXX.html" because that site is more responsive. Also ensured the ".html" suffix is there consistently.

Removed

  • The COLLECT_ANNOTATIONS and COLLECT_ERRORS options. Collection of these is now specified depending on whether a map is passed to the API to hold them.

Fixed

  • Annotations are now always collected in keywords that collect them instead of being skipped if the keyword does not validate.
  • Errors and annotations are now properly pruned, but at the schema level.
  • Fixed the starting "absoluteKeywordLocation" to be the root ID if it exists, otherwise it remains the base URI.

[0.12.0]

Added

  • Added the ability to add pluggable rules to the linter.
  • Added non-schema tree traversal to JSON.
  • JSONPath.endsWith for checking the last element.
  • New linter rule that checks for $ref elements having siblings, for Draft-07 and earlier.
  • New Id.element field that contains the ID's JSON parent element.
  • New Id.unresolvedID field that holds the unresolved ID URI.
  • ID scanning and validation is now performed on all known IDs and URLs before validation so that all references (to valid IDs) will resolve.
  • Added some total load and run times to the test runner.
  • All known schemas, including from URLs, are now validated.

Changed

  • Changed a couple linter behaviours:
    1. A string won't be examined if its parent is a definitions object, in addition to just a "properties" object.
    2. Unknown non-root keywords will have their properties examined. This is desirable because both definitions and "properties" objects allow unknown keywords. Previously, no unknown keywords had their properties examined.
  • Errors reverted to instance->schema order for the mapping.
  • All the fields are now final or private in Annotation and Id.
  • Other internal improvements.
  • Changed JSON.traverse to traverseSchema, JSON.JsonElementVisitor to JSON.SchemaVisitor, and JSON.TraverseState to JSON.SchemaTraverseState.
  • The Coverage tool now outputs two JSON objects: coverage by instance location and the seen-only schema locations.
  • Updated how JSON.traverseSchema works to be more complete. It now provides a more complete picture of the schema and the state of its members. This also means that all the keyword detection logic is in one place.
  • Updated the linter and coverage tool to utilize the new schema traverser features.
  • Options.set now returns itself, for easy chaining.
  • The "ipv4" and "ipv6" formats are now parsed using the internal URI parser. That uses a more current specification.
  • Changed LRUCache maximum size to be changeable.

Fixed

  • AUTO_RESOLVE now behaves properly for relative IDs.

[0.11.0]

Added

  • Added a rudimentary schema coverage tool, Coverage.

Changed

  • Introduced a JSONPath class that replaces strings for path representation. One of the advantages of this approach is not having to worry about "/" characters in names, or JSON Pointer syntax.
  • Other updates, optimizations, refactors, and improvements.
  • Errors are now mapped in schema->instance order instead of instance->schema order. This is now the reverse of how annotations are mapped.

Fixed

  • Strings.fromJSONPointerToken now throws an exception if the token contains a "/" character.
  • "error" annotations are now marked valid=true.
  • Annotations are now being properly collected for then and else.
  • Fixed URI.toString() and toDecodedString() by flipping their behaviour.
  • Fixed an absolute keyword location tracking bug.

[0.10.0]

Added

  • Added expected type checking to the linter. For example, minimum expects a type of "number" or "integer".
  • Added an "exclusive minimum" >= "exclusive maximum" check to the linter.
  • README updates: "Schema and instance coverage" future plans and "IDN Hostnames" link in the reference list.
  • Added annotation printing to Main.
  • Added an option, Option.COLLECT_ANNOTATIONS_FOR_FAILED, that controls whether annotation collection also retains annotations for schemas that failed validation.

Removed

  • The notes for Option.COLLECT_ERRORS stated that it may need to be false in the presence of any $recursiveRefs. This note has been removed.

[0.9.0]

Added

  • Added the source and javadoc package plugins to the POM.
  • A linter check for $id values that have an empty fragment, for Draft 2019-09 and later.
  • Added "minimum" keyword > "maximum" keyword checks to the linter.
  • Added a custom linting rule example to the README.

Changed

  • The EMAIL regex in Format now disallows local parts starting with a dot, ending with a dot, or containing two consecutive dots.
  • URI and hostname parsing is now down with local code in the com.qindesign.net package.
  • New internal URI and hostname parsing and handling. There's a new com.qindesign.net package containing URI and Hostname classes. URI replaces java.net.URI and Hostname handles both regular and IDN hostname parsing.
  • The linter now uses a list of strings for the path instead of a single JSON Pointer string.

Fixed

  • Improved the UUID format checker.
  • Javadoc updates and fixes.

[0.8.0]

Added

  • More idn-hostname format checks.
  • URL support in Main and Linter.
  • Internal $ref existence checking in Linter.
  • Table of contents in the README.
  • New "The linter" section in the README.

Fixed

  • There's now correct use of "an" vs. "a" in the error message in Type.
  • Annotation prefix checking now considers the final "/" and doesn't just match the prefix.

[0.7.1]

Changed

  • Changed the Maven artifactId to "snowy-json" from "json-schema".

[0.7.0]

Added

  • New section on annotations and errors in the README.
  • New annotation type descriptions in the README.
  • Optional test counts in the test runner.
  • New AUTO_RESOLVE option.
  • New Options.is method for reading Boolean values more conveniently.

Changed

  • Better and more consistent internal handling of anchors and IDs.

Removed

  • Annotation collection was removed from the Contains keyword.

Fixed

  • Fixed percent encoder, used when resolving absolute locations.
  • Relative paths resolved against a base URI having an empty path no longer result in an infinite loop.
  • Other code and Javadoc fixes.

[0.6.0]

Initial public release.

Copyright (c) 2020-2021 Shawn Silverman