fix(crush): fix handling of period-containing property names #95
Conversation
|
I have solved the "dot in property name" problem and added some more tests for the crush function. |
aleclarson
force-pushed
the
main
branch
2 times, most recently
from
2154f96 to
6a4b4f6
Compare
src/object/tests/crush.test.ts
Outdated
| const obj = { | ||
| 'a.b': { 'c.d': 123.4 } | ||
| } | ||
| expect(_.crush(obj)).toEqual({ | ||
| 'a.b.c.d': 123.4 | ||
| }) |
There was a problem hiding this comment.
This doesn't preserve the intent though. We'll probably need support for "a.b"."c.d" paths.
There was a problem hiding this comment.
@aleclarson I don't understand your remark.
The test is just a slightly more elaborate example of key names with a dot inside.
How does this not preserve intent ? It's just a test, it passes.
What do you mean with "a.b"."c.d" paths ? AFAIK it's not possible to specify such a path in javascript.
There was a problem hiding this comment.
Sorry for the vagueness. We can still merge this PR, since you fixed crush to pass the failed test case.
But in the future, we should avoid crushing ["a.b"]["c.d"] to a.b.c.d since that won't reproduce the same object when used with construct. Example below:
import { crush, construct } from 'radashi'
test('crush is the opposite of construct', () => {
const source = { 'a.b': { 'c.d': 1 } }
// This currently fails:
expect(construct(crush(source))).toEqual(source)
})So we can merge this PR now or, if you'd like, you could also fix the test case above and include that fix in this PR. Either way works, so just let me know!
There was a problem hiding this comment.
There may be purposes for crush() other than being the inverse of construct().
Probably a get() will be a good alternative for that in most cases, but the current crush() implementation is flawed as it returns undefined when there's dots in the propserty names..
I have made a new commit where the inner recursive function is outside of the crush() function, That should be slightly more efficient in term of closures.
There was a problem hiding this comment.
True. Okay, awaiting your new commit. 👍
There was a problem hiding this comment.
@aleclarson
I am unable to commit to my own PR ?
Get this error messages. Any idea what might be going on ?
$ git push
remote: Permission to radashi-org/radashi.git denied to stefaanv.
fatal: unable to access 'https://github.com/radashi-org/radashi.git/': The requested URL returned error: 403
There was a problem hiding this comment.
I don't know what your git remote -v looks like, but seems like you need to set the branch's upstream with:
git remote set-url origin https://github.com/yourusername/radashi.git
git push -u origin your-branch-name
Replace 'yourusername' with your GitHub username and 'your-branch-name' with your actual branch name. This will update your origin to point to your fork and then push your branch there.
After this, you can just use git push for future pushes on this branch.
There was a problem hiding this comment.
@aleclarson thanks for the help, I succeeded in committing this changes.
You can merge the PR
aleclarson
force-pushed
the
fix/crush-dot-property
branch
from
0a7bfb9 to
b7370c7
Compare
There was a problem hiding this comment.
Thank you for spending time on this!
I've reworked it, but the general solution you implemented remains intact. I added a failing test case, which I then fixed, that will ensure support for built-in object types like Map, Set, etc (these will not be crushed).
I also took time to improve the return type. If a crushed object has nested objects, we cannot assume anything about the value types, because of TypeScript limitations. In particular, we cannot know which object types will pass the isObject check (and therefore get crushed). There are “halfway solutions” to this, like keeping a union type of all the built-in object types, but I opted for the safer approach of “assume nothing”.
Let me know what you think! And thanks again.
aleclarson
changed the title
|
Oh btw, I undid your change in b3eef45, as hoisting functions is a micro-optimization that Radash (and by extension, Radashi) hasn't done in the past. If you have a strong opinion on this, we can discuss over in the discussions area. 👍 |
|
@aleclarson |
|
Sorry, I guess I've pushed the close button too soon |
|
All merged now. Just wanted to wait in case you had any objections :) |
|
A new beta version To install: pnpm add [email protected]The 
|
Resolves sodiray/radash#365
Resolves #23