Sheet Management

In the Graviti SDK, Sheet and DataFrame are interpretations of the same thing at different levels.

Sheet refers to the form of data organization that is one level lower than the dataset. One dataset can have many different sheets, such as train, test, or frame-by-frame pictures from different videos. Each sheet has its own schema.

Graviti SDK organizes the data of a sheet into a DataFrame format, which makes it more convenient and intuitive to get and modify the data.

More details about the DataFrame, Binary Files, Schema and Search are as follows:

• DataFrame
• Binary Files
• Schema
• Search

In the Graviti SDK, the relationship between sheet name, DataFrame and dataset/draft is like the relationship between key, value and dict. Thus, SDK supports managing sheets like manipulating the dict in python.

The following will introduce more details about the sheet management methods in the SDK. First of all, it is necessary to get a dataset:

from graviti import Workspace

ws = Workspace(f"{YOUR_ACCESSKEY}")
dataset = ws.datasets.get(f"{DATASET_NAME}")

Create a Sheet

SDK supports managing sheets of open drafts. Thus, it is necessary to create or get an open draft:

draft = dataset.drafts.create(f"{DRAFT_TITLE}")

Then users need to Initialize a DataFrame like:

>>> from graviti import DataFrame
>>> data = [
...    {"filename": "a.jpg"},
...    {"filename": "b.jpg"},
...    {"filename": "c.jpg"},
... ]
>>> df1 = DataFrame(data)
>>> df1
   filename
0  a.jpg
1  b.jpg
2  c.jpg

Next, users can create a new sheet:

draft[f"{SHEET_NAME}"] = df1
draft.upload()

Note

Only changes made on the open draft can be synchronized to Graviti via the method upload().

Or replacing the old sheet:

draft["train"] = df2
draft.upload()

List Sheets

List sheets for the specified version of the dataset:

dataset.checkout(f"{COMMIT_ID}")
list(dataset.keys())

List sheets of an open draft:

draft = dataset.drafts.get({DRAFT_NUMBER})
list(draft.keys())

Get a Sheet

Get a sheet for the specified version of the dataset by name:

dataset.checkout(f"{REVISION}")
dataset[f"{SHEET_NAME}"]

Get a sheet of an open draft by name:

draft = dataset.drafts.get({DRAFT_NUMBER})
draft[f"{SHEET_NAME}"]

Delete a Sheet

Delete a sheet of the open draft by name:

del draft[f"{SHEET_NAME}"]
draft.upload()