Feature/type search #47
Conversation
|
|
||
| response_object = AethelListResponse() | ||
|
|
||
| # First we select all relevant samples from the dataset that contain the query string. | ||
| query_result = search( |
There was a problem hiding this comment.
It turns out that this pre-filtering loop is not necessary, and we can improve performance slightly by removing it. I compared word and type searches with and without this step using a bigger dataset (3000 samples). Word searches performed slightly better without the pre-filtering (0.04s) step than with it (0.05s), and the same is true for type searches (0.52s with pre-filtering => 0.42s without).
There was a problem hiding this comment.
As an added benefit, we can drop a lot of code taken from aethel in search.py.
|
Performance is still an issue when the full dataset is used. The last commit provides some improvements regarding the retrieval of existing results in the response object, but further optimisations are needed. Two possible optimisations in the frontend will be taken up in separate PRs for these issues: |
This PR closes #37, implementing type-based searches in Aethel. It also implements #39 in a basic way for testing purposes. The styling of the buttons could use some work, but at least the functionality is there.