Slow dataset queries

[forgetso]

We are using $sample 2 when getting captchas. This is causing slow queries on the nodes. We need to change this approach as follows:

1. Create an index on { datasetId: 1, solved: 1 }

2. Instead of $sample, use a random selection method to improve performance. For example:

• Add a random field to each document at insertion time.
• Index this field.
• Query using $gte or $lte to efficiently retrieve random documents.

3. Use $limit Before $sample

Instead of sampling from the entire dataset, limit the query first:

db.captchas.aggregate([
  { $match: { datasetId: "0xe666b35451f302b9fccfbe783b1de9a6a4420b840abed071931d68a9ccc1c21d", solved: true } },
  { $limit: 1000 },  // Get a subset first
  { $sample: { size: 2 } },  // Then sample from that subset
  { $project: { datasetId: 1, datasetContentId: 1, captchaId: 1, captchaContentId: 1, items: 1, target: 1 } }
]);

This reduces the number of documents MongoDB has to scan.

[goastler]

aggregate has no ordering so you don't need the random field

[forgetso]

https://github.com/prosopo/captcha/pull/1705