Add exact param in remove doc by word method #41
Conversation
|
This is perfect, thank you @mateonunez 🙏 |
| const id = search.hits[0].id; | ||
| await db.delete(id); | ||
|
|
||
| const search2 = await db.search({ term: "abc" }); |
There was a problem hiding this comment.
This line should also use exact: true, otherwise the tests passes even if #40 is not fixed.
There was a problem hiding this comment.
If you search for abc as an exact term, the result will be 0, 'cause the only document available is the document that contains txt: "abcd".
This expected result could be more realistic, but in both cases, it works correctly
There was a problem hiding this comment.
But it shouldn't be 0, that's the bug in #40. It should be 1, because there is a second document containing "abc" that should not be deleted.
The test adds two documents containing "abc" and then deletes one of them. The other one should still be there, and not be deleted.
At a glance (but I did not spend much time delving into the code, so I might be wrong), the issue is that the term should not be deleted unless all the documents associated to it were deleted. Instead, the code sets node.end = true, effectively removing the term, even if it still has documents associated to it.
There was a problem hiding this comment.
Got the point. The issue should be opened again.
| const root = this.root; | ||
| if (!word) return false; | ||
|
|
||
| function removeWord(node: TrieNode, _word: string, docID: string): boolean { | ||
| const [nodeWord /**_docs*/] = node.getWord(); | ||
|
|
||
| if (node.end && nodeWord === word) { | ||
| if (node.end || (exact && node.end && nodeWord === word)) { |
There was a problem hiding this comment.
#38 cannot be caused by that, because it occurs even if no document is deleted, so this code path is never called in the reproduction.
|
|
||
| const serach2 = await db.search({ term: "stelle" }); | ||
|
|
||
| expect(serach2.count).toBe(1); |
There was a problem hiding this comment.
This assertion should check that the search hits do not include the deleted ID, not that the count is 1.
This PR fixes #39 and #40.
I've implemented a new param in the
removeDocByWordmethod used inremovemethod to prevent removing words that not exactly matches the document's word.I also added 2 cases to the tests.