tl;dr: is there a well defined process for amending the current LSST alert schema or creating alerts with a user-defined schema?
As part of the PLAsTiCC v2.0 broker challenge, we’d like to supplement the standard alert schema with host information. I’ve generated a notebook to add additional fields, but wanted to know the best practice for writing out these schema.
Using schema = lsst.alert.packet.Schema.from_file(path/to/our/modified/schema/lsst.v4_1.alert.avsc)
seems to throw up exceptions that it is still expecting a valid v4_0 defined schema.
Note that the modified schema are here: plasticc_alerts/Examples/plasticc_schema at main · LSSTDESC/plasticc_alerts · GitHub (the same workbook contains our work).
Hi @reneehlozek, it looks to me like some of the namespace lines in your 4.1 schemas still specify the 4.0 schema (e.g., here)–I would see if changing these to 4.1 would help.
I will mention/warn that we (@spencernelson) are considering some larger changes to alert_packet to make it simpler: any eventual proposal will show up at DM-28733.
Setting aside the technical question: Rubin is not going to have these host quantities available in our alerts (because the information is unknown!)–does it set the wrong expectation for PLAsTiCC data consumers to include them?
hi @ebellm thanks for the input and technical help, that definitely fixed it. I’m excited to see the proposal on changes too, so thanks for the heads up. I agree that this exhaustive list is beyond the scope of what Rubin can provide (a more sensible list would be host mags + errors + light profile moments + PSF), but this will hopefully be a jumping off point for discussions about the alerts and how folks will use ancillary data in classifications.