Skip to content

Configuration with Header

For a CSV file with header, you need to set withHeader to true in the configuration file, indicating that the first behavior in the CSV file is the header. The header content has special meanings.


If the CSV file contains headers, the Importer will parse the Schema of each row of data according to the headers and ignore the vertex or edge settings in the YAML file.

Sample files

The following is an example of a CSV file with header:

  • sample of vertex

    Example data for student_with_header.csv:


    The first column is the vertex ID, followed by the properties name, age, and gender.

  • sample of edge

    Example data for follow_with_header.csv:


    The first two columns are the start vertex ID and destination vertex ID, respectively. The third column is rank, and the fourth column is property degree.

Header format description

The header defines the start vertex, the destination vertex, the rank, and some special functions by keywords as follows:

  • :VID(mandatory): Vertex ID. Need to use :VID(type) form to set data type, for example :VID(string) or :VID(int).
  • :SRC_VID(mandatory): The start vertex ID of the edge. The data type needs to be set in the form :SRC_VID(type).
  • :DST_VID(mandatory): The destination vertex ID of the edge. The data type needs to be set in the form :DST_VID(type).
  • :RANK(optional): The rank value of the edge.
  • :IGNORE(optional): Ignore this column when inserting data.
  • :LABEL(optional): Insert (+) or delete (-) the row. Must be column 1. For example:



All columns except the :LABEL column can be sorted in any order, so for larger CSV files, the user has the flexibility to set the header to select the desired column.

For Tag or Edge type properties, the format is <tag_name/edge_name>.<prop_name>:<prop_type>, described as follows:

  • <tag_name/edge_name>: Tag or Edge type name.
  • <prop_name>: property name.
  • <prop_type>: property type. Support bool, int, float, double, timestamp and string, default string.

Such as,

Sample configuration

version: v2

description: example

# Whether to delete temporarily generated logs and error data files.
removeTempFiles: false


  # Retry times of nGQL statement execution failures.
  retry: 3

  # Number of NebulaGraph client concurrency.
  concurrency: 10 

  # Cache queue size per NebulaGraph client.
  channelBufferSize: 128

  # Specifies the NebulaGraph space to import the data into.
  space: student

  # Connection information.
    user: root
    password: nebula
    address: 192.168.*.13:9669

    # Configure some of the operations to perform after connecting to the NebulaGraph server, and before inserting data.
    commands: |
      DROP SPACE IF EXISTS student;
      CREATE SPACE IF NOT EXISTS student(partition_num=5, replica_factor=1, vid_type=FIXED_STRING(20));
      USE student;
      CREATE TAG student(name string, age int,gender string);
      CREATE EDGE follow(degree int);

    # The interval between the execution of the above command and the execution of the insert data command.
    afterPeriod: 15s

    # Configure some of the actions you performed before disconnecting from the NebulaGraph server.
    commands: |

# Path of the error log file.
logPath: ./err/test.log

# CSV file Settings.

    # Path for storing data files. If a relative path is used, the path is merged with the current configuration file directory. The first data file in this example is vertex data.
  - path: ./student_with_header.csv

    # Insert the failed data file storage path, so that data can be written later.
    failDataPath: ./err/studenterr

    # The number of statements inserting data in a batch.
    batchSize: 10

    # Limit on the number of rows of read data.
    limit: 10

    # Whether to insert rows in the file in order. If the value is set to false, the import rate decreases due to data skew.
    inOrder: true

    # File type. Currently, only CSV files are supported.
    type: csv

      # Whether there is a header.
      withHeader: true

      # Whether there is a LABEL.
      withLabel: false

      # Specifies the delimiter for the CSV file. A string delimiter that supports only one character.
      delimiter: ","

      # Schema type. Possible values are vertex and edge.
      type: vertex

    # The second data file in this example is edge data.
  - path: ./follow_with_header.csv
    failDataPath: ./err/followerr
    batchSize: 10
    limit: 10
    inOrder: true
    type: csv
      withHeader: true
      withLabel: false
      # The type of Schema is edge.
      type: edge
        # Edge type name.
        name: follow

        # Whether to include rank.
        withRanking: true


The data type of the vertex ID must be the same as the data type of the statement in clientSettings.postStart.commands that creates the graph space.

Last update: February 19, 2024