/v2/updates/price/timestamp returns "Update data not found" regardless of ignore_invalid_price_ids flag.

[comunidadio]

The API returns HTTP 404 "Update data not found" whenever one update is not found for any of the price feed IDs requested.
It's cumbersome as there is also no indication of which of the passed feed IDs did not have an update available at that time (so request could be retried without that feed ID).

The failure is triggered at:

pyth-crosschain/apps/hermes/server/src/api/rest/v2/timestamp_price_updates.rs

Line 109 in 2faddf9

RestError::UpdateDataNotFound

Expecting that at least when ignore_invalid_price_ids=true, Hermes would return whatever updates it had available and let client handle this.

[aditya520]

Can you give us the script/code you are running?

[comunidadio]

Apologies for the delay coming back to this.

It's easily reproducible by taking feed IDs from /price_feeds endpoint then asking a price update:

Example:
https://hermes.pyth.network/v2/updates/price/1759881331?encoding=base64&parsed=true&ignore_invalid_price_ids=true&ids[]=f7ac0d8fdf22640d39de13816cd0942aa6cf3874b95028d9a602056e1dc95e96&ids[]=51ea670d2a41ee695d8c4b4722f6f45cdfbdfcdb42eed8c03e11596cd0db5444
returns "update data not found".
Here 51ea670... looks like a test feed (but that's not programatically easy to differentiate from /price_feeds as we have legit "TST/USD TEST Token" 😅), but it also ends up the same with any feed that did not have update at that timestamp (eg. markets that are closed like FX on weekends or stocks).

Since we're ok with ignore_invalid_price_ids, ideally the query should just return the data it has and let the consumer handle this.

[sonicd300]

Im having the same problem while querying some of the official contract ids returned by /price_feeds although it seems weird, but isnt that the point of having the ignore flag set to true?

[aditya520]

@tejasbadadare Can you have a look?

[tejasbadadare]

Hey @comunidadio , the ignore_invalid_price_ids flag is meant to ease migration troubles for users that have subscribed to a set of feeds, and one or more feeds later become deprecated. The flag allows the old subscription request to still work and just filters out the unrecognized feed IDs. The RestError::UpdateDataNotFound error is a bit different, it captures the case where the feed exists, but there is no valid update that exists at the exact timestamp.

Because the feed is timeseries data, it's not guaranteed that each timestamp will have a price -- you mentioned some cases such as market closed. We define the canonical price at timestamp T to be the first price where publish_time >= T.

However, you bring up a good point that it's cumbersome to identify which of the feeds caused the error, and it would be friendlier to just return the feeds that do have a price at that timestamp. We would likely add this under a different flag since it's different behavior than what ignore_invalid_price_ids is intended to capture as mentioned above. I'll circulate this with the team and let you know how we prioritize it. Thank you!