You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In #4557 we're adding a new type that concisely encodes the set of features used by a sstable. For now, this is derived from the Properties when we open a sstable. Ideally in the future we wouldn't need to read the properties block at all when we open a sstable.
When we open a sstable, the first thing we do is read the fixed-length footer at the very end of the file. See this comment for more context on the layout of a sstable. We should begin writing the uint32 feature set as a part of the footer. This is a backwards incompatible change, so we'll need to introduce a new TableFormat.
At write time we should compute the feature set from the properties we've derived, and then include it in the footer we write out (but only if the configured table format is sufficiently high).
At read time after we parse the table format, we'll extract the feature set only if the table format is sufficiently high. For now we can just plumb it up and then validate that the feature set we derive from the properties block matches the feature set we read from the footer.
Eventually we'll be able to avoid reading the properties block when constructing a new reader if the table has a sufficiently high table format.
In #4557 we're adding a new type that concisely encodes the set of features used by a sstable. For now, this is derived from the
Properties
when we open a sstable. Ideally in the future we wouldn't need to read the properties block at all when we open a sstable.When we open a sstable, the first thing we do is read the fixed-length footer at the very end of the file. See this comment for more context on the layout of a sstable. We should begin writing the
uint32
feature set as a part of the footer. This is a backwards incompatible change, so we'll need to introduce a new TableFormat.At write time we should compute the feature set from the properties we've derived, and then include it in the footer we write out (but only if the configured table format is sufficiently high).
At read time after we parse the table format, we'll extract the feature set only if the table format is sufficiently high. For now we can just plumb it up and then validate that the feature set we derive from the properties block matches the feature set we read from the footer.
Eventually we'll be able to avoid reading the properties block when constructing a new reader if the table has a sufficiently high table format.
Jira issue: PEBBLE-429
Jira issue: PEBBLE-435
The text was updated successfully, but these errors were encountered: