Open evansd opened 3 months ago
This could be solved using the same approach as used here:
We'd need to update the code which gets information about the schemas to return an is_raw
flag for each schema:
https://github.com/opensafely-core/ehrql/blob/f5b0d5f56b53039062cf1f95ea76dda584f485de/ehrql/docs/schemas.py#L43-L52
Probably this would just be based on the module name as I don't think we have anything more structured to use here.
Then the code which renders the schemas to Markdown would need to add a metadata header to the page which configures the search boost appropriately: https://github.com/opensafely-core/ehrql/blob/f5b0d5f56b53039062cf1f95ea76dda584f485de/ehrql/docs/render_includes/schemas.py#L40-L63
For raw pages I imagine something like this would do the trick, but we'll need to experiment with the boost values a bit:
---
search:
boost: 0.001
---
Searching for tables which appear in both "cooked" and "raw" schemas returns the raw schema first, which confuses users.
For example:
Thread where this confusion arose: https://bennettoxford.slack.com/archives/C04DVD1UQC9/p1709896596557359