Long-term we want that migration to be done entirely in SQL, but for now this is a hotfix to not cause OOMs on large databases. This is using a homegrown version of `Repo.stream`, it's worse in terms of performance than the upstream since it doesn't use the same prepared query for chunk queries, but unlike the upstream it supports preloads. |
||
|---|---|---|
| .. | ||
| mix | ||
| pleroma | ||
| jason_types.ex | ||
| transports.ex | ||
| xml_builder.ex | ||