You are viewing a plain text version of this content. The canonical link for it is here.

Posted to reviews@iotdb.apache.org by GitBox <gi...@apache.org> on 2021/11/14 09:43:15 UTC

[GitHub] [iotdb] JackieTien97 opened a new pull request #4382: New architecture of aligned timeseries

JackieTien97 opened a new pull request #4382:
URL: https://github.com/apache/iotdb/pull/4382

## Description

Previously, we use `vector` to contains a batch of timeseries who share the same time column. In previous implementation, we treat `vector` as a special `MNode` in `MTree` which is counter-intuitive conceptually.

In this PR, we treat aligned as an attribute of device. That means a batch of timeseries who share the same time column is grouped under a device and then mark that device as aligned device. So, there is no more any special `MNode` in `MTree`. Instead, we add a new field `isAligned` in `EntityMNode`.

As device has `isAligned` attribute, so the timeseries under it are either all aligned(share the same time column) or non-aligned(each timeseries has its own time column). It will never happen that one device has both aligned and non-aligned timeseries.

Previously, we cannot add or delete sub sensors in `vector`. Currently, you can add or delete any aligned timeseries under that device as you like.

Previously, you can only delete from data of all sub sensors for one `vector` in some time range. Currently, you can treat aligned timeseries as common timeseries while doing deletion. You can delete from one of aligned timeseries under one device.

In previous implementation, in you create `s1`, `s2` and `s3` as vector, even if you only write `s1` and `s2` for a certain period of time, we will waste some space in disk to save an empty `chunk` or `page` for `s3`. In this pr, we won't do this, we only save aligned timeseries which have data. Those aligned timeseries which don't have data, we won't even waste one bit for them.

We also add `MeasurementPath` and `AlignedPath` to represent for different types of full path. In previous implementation, we only have `PartialPath` which could stand for both partial path and full path. However, in some code path, like write and query executors, these `PartialPath` should all be fullPath. And some attributes like `alias` which should only exist in full path are removed from `PartialPath` to `MeasurementPath` and `AlignedPath`.

After we add `MeasurementPath` and `AlignedPath` to represent full path, we could put schema info into these two class. In previous implementation, while parsing query sql to logical plan and then to physical plan, we first get `PartialPath` from `MTree` and then use thse `PartialPath` to ask `Mtree` again for schema info like data types which is redunctant. Now, after `removeStars` methods, we get full path, the schema info in already packed into path, we can directly use it in the following functions.

There are many kinds of query in current IoTDB, so we will not support all of these for aligned timeseries in this PR., but it will be supported in the future. The following kinds of query are supported in this pr:
- [*] raw query without value filter
- [*] raw query without value filter

--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: reviews-unsubscribe@iotdb.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org