Diff

TensorBay supports showing changes between commits or drafts.

Before operating the diff, a dataset client instance with commits is needed. See more details in Draft and Commit

Get Diff

TensorBay SDK allows getting the dataset diff through basehead. Currently, only obtaining the diff between the head and its parent commit is supported; that is, the head is the given version(commit or draft) while the base is parent commit of the head.

diff = dataset_client.get_diff(head=head)

The type of the head indicates the version status: string for commit, int for draft.

Get Diff on Revision

For example, the following diff records the difference between the commit whose id is "3bc35d806e0347d08fc23564b82737dc" and its parent commit.

diff = dataset_client.get_diff(head="3bc35d806e0347d08fc23564b82737dc")

Get Diff on Draft Number

For example, the following diff records the difference between the draft whose draft number is 1 and its parent commit.

diff = dataset_client.get_diff(head=1)

Diff Object

The structure of the returning DatasetDiff looks like:

dataset_diff
├── segment_diff
│   ├── action
│   │   └── <str>
│   ├── data_diff
│   │   ├── file_diff
│   │   │   └── action
│   │   │       └── <str>
│   │   └── label_diff
│   │       └── action
│   │           └── <str>
│   └── ...
├── segment_diff
│   ├── action
│   │   └── <str>
│   ├── data_diff
│   │   ├── file_diff
│   │   │   └── action
│   │   │       └── <str>
│   │   └── label_diff
│   │       └── action
│   │           └── <str>
│   └── ...
└── ...

The DatasetDiff is a list which is composed of SegmentDiff recording the changes of the segment. The SegmentDiff is a lazy-load sequence which is composed of DataDiff recording the changes of data.

The attribute “action” represents the status difference of the relative resource. It is an enum which includes:

• unmodify

• add

• delete

• modify