Conversation
|
|
||
| - pco supports `u16`, `u32`, `u64`, `i16`, `i32`, `i64`, `f16`, `f32`, `f64` | ||
| - pco_store adds support for `SystemTime`, `bool` | ||
| - pco_store adds support for `SystemTime` (mapped to ???), `bool` (mapped to ???) |
There was a problem hiding this comment.
- pco_store adds support for
SystemTime(stored asi64),bool(stored asu16)
| ``` | ||
|
|
||
| `STORAGE EXTERNAL` is set so that Postgres doesn't try to compress the already-compressed fields | ||
| The pco-compressed columns are expected to typically be in Postgres [TOAST](https://www.postgresql.org/docs/current/storage-toast.html). Using `STORAGE EXTERNAL` is recommended so that Postgres doesn't try to compress the already-compressed fields, speeding up writes. |
There was a problem hiding this comment.
I think the TOAST sentence should be removed because it's just explaining how Postgres works, and pco_store would work whether or not TOAST existed in its current form.
I'd also reword the existing sentence:
STORAGE EXTERNALis recommended to improve write times by skipping the default Postgres compression step.
| The pco-compressed columns are expected to typically be in Postgres [TOAST](https://www.postgresql.org/docs/current/storage-toast.html). Using `STORAGE EXTERNAL` is recommended so that Postgres doesn't try to compress the already-compressed fields, speeding up writes. | ||
|
|
||
| This uses a `(end_at, start_at)` index because it's more selective than `(start_at, end_at)` for common use cases. For example when loading the last week of stats, the `end_at` filter is what's doing the work to filter out rows. | ||
| Its recommended to index `(end_at, start_at)`, because it's more selective than `(start_at, end_at)` for common use cases. For example when loading the last week of stats, the `end_at` filter is what's doing the work to filter out rows. |
There was a problem hiding this comment.
Typo: It's. I'm also not sure the added comma makes sense
| ## Performance | ||
|
|
||
| Numeric compression algorithms take advantage of the mathematic relationships between a series of numbers to compress them to a higher degree than binary compression can. Of the numeric compression algorithms available in Rust, pco achieves both the best compression ratio and the best round-trip read and write time. | ||
| Numeric compression algorithms take advantage of the mathematic relationships between a series of numbers to compress them to a higher degree than binary compression can. Of the numeric compression algorithms available in Rust, in our tests pco achieves both the best compression ratio and the best round-trip read and write time. |
There was a problem hiding this comment.
This might read better:
In our tests of the numeric compression algorithms available in Rust, pco achieves both the best compression ratio and the best round-trip read and write time
|
Looks like you'll need to fix the commit history; after the other PR was merged its commit started showing up in the diff here |
No description provided.