All notable changes to this project will be documented in this file. See Conventional Commits for commit guidelines.
- Minimum supported Node version is v14.18; we're dropping v12 support
- export types (11b5fb9)
- tweak types (4791f68)
- export defaults and version (1107244)
- migrate to ES Modules (8c9d95d)
- programs now are in ES Modules and won't work with Common JS
require()
- config file based major bump blacklisting (e15f9bb)
- Revert "chore: setup refresh" (23cf206)
- some date object support and corrections for the callback values (9017893)
- no changes in API; only correction to the callbacks; things now work as intended in callbacks; bumping just in case
- nothing should be breaking, there are no API changes, but we corrected the callback values, correct types and correct original pair values are passed. Before, the first value passed to the callback might have been not the original one, it might have been pre-merged.
- tackle date objects - they're not part of JSON spec - but if two values are date objects, newer date will prevail
- add
testStatsto npmignore (f3c84e9)
- recode in TS and named exports (5981d20)
- previously it was consumed like:
import mergeAdvanced from ...- now like:import { mergeAdvanced } from ...
Accidental version bump during migration to SourceHut. Sorry about that.
- remove options checking to make program run around 128 times faster (9ce7714)
- tap
Array.isArraydirectly, gaining around 124 times more speed (13cd986)
- Various documentation and setup tweaks after we migrated to monorepo
- Setup refresh: updated dependencies and all config files using automated tools
- Added the fourth argument in the callback (306e3da)
- Updated all dependencies
- Restored coveralls.io reporting
- Restored unit test linting
- Set up Rollup to remove comments from all builds
- Removed AVA ESLint rules until the plugin is fixed
GitHub sold us out. In the meantime, we:
- Migrated to BitBucket (to host repo + perform CI) and Codacy (for code quality audit)
- Dropped BitHound (RIP) and Travis
- Removed
package-lock
- Fourth argument in the callback. It will allow to make decisions evaluating key names, paths and their type (is it a key of an array or an object?).
- Set up Prettier
- Removed
package.lockand.editorconfig - Wired Rollup to remove comments from non-dev builds. This means we can now leave the
console.logs in the source code — Rollup will remove from production code. - Unit tests are pointing at ES modules build, which means that code coverage is correct now, without Babel functions being missed
- Dropped
airbnb-baseas ESLint preset and set up Prettier. Now contributors don't have to worry about the code style any more.
- When
opts.useNullAsExplicitFalseis on and one of the clashing values isnull, result will benull, notfalse. That's a breaking change of the API and this warrants a major semver bump.
PS. This feature is needed to maintain the data integrity. When merging in multiple rounds, if null under opts.useNullAsExplicitFalse yielded false, after the first round, the null would be lost. Now, null is kept and all subsequent merges will yield null. Practically, this means that it's enough to place null anywhere on any any level of template data and the result is guaranteed to be null. It's super-easy way to remove default values arrays or objects — the merged result null will not cause defaults to show up now.
opts.cb- Rebased a lot of code, merged
util.jsinto the main file - Code coverage is now 100% line & function-wise
Previously, opts.hardMergeKeys, opts.ignoreKeys as well as opts.hardMergeEverything and opts.ignoreEverything were not 100% unidirectional. In some cases, for example, when object key's value was an array and it clashed with another array from same-named key, the merging was performed instead of hard write. It was done with intention to retain as much data as possible after merging. However, it was not what was promised in documentation. Now this is sorted.
Also, single asterisk wildcard * among opts.ignoreKeys sets opts.ignoreEverything to true. Same with * among opts.hardMergeKeys.
Thanks to @jabiinfante PR:
- Add
opts.hardArrayConcat - Add
opts.hardArrayConcatKeys
- Add
opts.dedupeStringsInArrayValuesnow returns sorted arrays
- Add
opts.useNullAsExplicitFalse
- Add
opts.mergeBoolsUsingOrNotAnd
- Recoded everything in ES modules, set up Rollup and now generating 3 flavours: CommonJS, UMD and ES modules.
opts.concatInsteadOfMergingopts.dedupeStringsInArrayValues
PS. Bumping major just in case.
- Switched from JS Standard to ESLint on
airbnb-basepreset with override to ban semicolons. 3 reasons: 1. JS Standard uses outdated ESLint which hasshell.jsone level-deep which is insecure. 2. JS Standard cares pretty much only about semicolons and doesn't activate many useful rules thatairbnb-*have on. 3. Using raw ESLint guarantees, well, latest ESLint, always. - Fixed many style things that ESLint flagged up. Functionality stays the same.
- Now serving transpiled code. Sorry everybody impacted by this, I got onto Sindre's bandwagon but I guess it's too early. I'll transpile all my libs.
- Tweaked bithound config.
Bumping the major just in case because strictly speaking, while things behave more correct now, they behave slightly different. This involves more precise hard merging and ignoring keys, when values are container-like (arrays or plain objects). For the rest, the merging algorithm's behaviour is the same. Read below.
- Improvements to the algorithm, related to hard merges and ignores. Quick refresher: a hard merge is when second item overwrites first no matter what. An ignore is the opposite - first value is left as it is, no matter what it is merged against. Up until now, the hard merging was done via simple overwriting. On some edge cases I found out that container-like values (arrays or objects) should not be simply overwritten, but their contents should be hard-merged instead. Again, this concerns only narrow cases where types of both merger and mergee match and are equal to arrays or plain objects.
- Add
opts.oneToManyArrayObjectMerge. It's an essential feature when setting defaults on data structures in JSON. When your default values object has array with single object, but your working object has an array with many objects. That's one-to-many merge.opts.oneToManyArrayObjectMergeis off by default in order to keep the release semver "minor". 🦄
- 🔧 Slightly rebased so that there's 100% branch coverage too, not only statements/lines.
- Switched to check-types-mini; removed all existing functions responsible for options' types.
- Add
opts.mergeArraysContainingStringsToBeEmpty
- When the options object values are set to be of a wrong type, this library will throw an error. At the same time, we're trying to be as flexible as possible - for example, accepting single string value instead of array.
- new options setting,
opts.ignoreKeys- this is one direction merge. Key names accept wildcards. - new options setting,
opts.hardMergeKeys- this is also one direction (opposive of above) merge. Key names accept wildcards.
It took nearly whole weekend on and off to get this out. Bloody leading wildcard bug kept poking out until I switched to matcher, problem solved.
It's nice to use the modular way of coding JS: I found bug here on this library with leading wildcards, but it was coming from dependency, another of mine, array-includes-with-glob, which in turn was bugged by buggy wildstring. I swapped its dependencies, updated object-merge-advanced (this library) and tests passed again.
Like changing a broken carburettor with another, working-one. 🎉
- 🔧 Recoded everything, adding Number type. This means, now there are 100 possibilities of the merge. Looks even more clean and optimal now, but that's a subjective thing.
- Added an optional options object, the third argument. Now
opts.mergeObjectsOnlyWhenKeysetMatchesallow more granular control over how objects within arrays are merged. The default setting is balanced option,opts.mergeObjectsOnlyWhenKeysetMatches = falseis more gung-ho merging approach (behaviour like v4, previous version).
- 🔧 Updated dependencies, now requesting JS Standard as normal version range, to prevent future surprises.
- 🔧 Improved the algorithm, removed redundant insurance, cloning input variable.
-
🔧 Making API even more user-friendly. If one of the input args is missing, instead of returning the
undefinednow it's returning the argument that's present. What's the point to returnundefined?If none are present,
undefinedis returned as before.
- 🔧 Technically a major API change. When object key values are arrays now we're checking is merger's element already present in a mergee. If so, it's omitted. This means, you can safely merge similar arrays without them bloating. However, in theory, while it's very logical and necessary feature, it's also a major API change. Hence bumping to v.3.
By the way, I needed this myself, placeholder default values in merged JSON files otherwise get duplicated. Now it's how it should be. 🍺
- 🔧 Major API change. Input argument objects are not mutated any more. Function first clones what it later uses.
- 🔧 Adding tests for input argument mutation (
3.xgroup). - 🔧 All auxiliary functions are ported inside the main exported function. Looks cleaner.