[BUG] QuerySearch Value error even though the shape of inputs is correct

[PavelRechkalov]

1. I fit the QuerySearch on a tensor of shape (561,1,4)
2. Then I provide it a query of shape (1,4)
3. I get the following error message:

      """
      ValueError: The length of the query should be inferior or equal to the length of data (X_)
      provided during fit, but got 4 for X and 4 for X_
      """

I use the latest aeon version in python 3.12.3 env

Kind regards

[PavelRechkalov]

You can reproduce this error even using the datasets provided by the library:

from aeon.datasets import load_classification
X, y = load_classification("GunPoint")
X_train, q =  X[:-1,:,:],  X[-1,:,:]

querySearcher = QuerySearch(
    k = 3
   ,distance = 'euclidean'
   ,normalise = False
   ,speed_up = 'fastest'
   ,n_jobs = 1 #os.cpu_count() - 1
) 
querySearcher.fit(X = X_train)
querySearcher.predict(q)

[PavelRechkalov]

I believe the issue might be the marked in red IF condition:

Should not it be ">" instead of ">="?

[baraline]

Hi, thanks for raising the issue, FYI most of this code is getting reworked in #2473, and will contain breaking changes.

Concerning the current issue you are facing, you are right, the if condition here should be if query_length > self.min_timepoints_. I'll check in #2473, but I don't think the issue appears in the new code.
For fixing it on your side, I would recommend editing the source file for now.

Note that having query_length == self.min_timepoints_ is equivalent to doing a whole series KNN, rather than working on subsequence similarity, and hence the computational optimisations will not help much !

[PavelRechkalov]

Greetings sir, thank you for replying.

Indeed, I updated the code on my end, now it works as intended.

When the #2743 is expected to be released? I'm asking to adjust my development process

Thank you in advance

[baraline]

When another maintainer have time to review it, it's quite a big one. Furthermore, for now, some functionalities have been removed (for example, query search on a collection of series) I would need to make some wrappers to make that happens, but this is simple. If you need any assistance once the PR is in to adapt your code, don't hesitate to create a discussion!