pymovements.gaze.from_csv#

Initialize a pymovements.gaze.gaze_dataframe.GazeDataFrame.

Parameters:

file (str | Path) – Path of gaze file.
experiment (Experiment | None) – The experiment definition. (default: None)
trial_columns (list[str] | None) – The name of the trial columns in the input data frame. If the list is empty or None, the input data frame is assumed to contain only one trial. If the list is not empty, the input data frame is assumed to contain multiple trials and the transformation methods will be applied to each trial separately. (default: None)
time_column (str | None) – The name of the timestamp column in the input data frame. (default: None)
time_unit (str | None) – The unit of the timestamps in the timespamp column in the input data frame. Supported units are ‘s’ for seconds, ‘ms’ for milliseconds and ‘step’ for steps. If the unit is ‘step’ the experiment definition must be specified. All timestamps will be converted to milliseconds. (default: ‘ms’)
pixel_columns (list[str] | None) – The name of the pixel position columns in the input data frame. These columns will be nested into the column pixel. If the list is empty or None, the nested pixel column will not be created. (default: None)
position_columns (list[str] | None) – The name of the dva position columns in the input data frame. These columns will be nested into the column position. If the list is empty or None, the nested position column will not be created. (default: None)
velocity_columns (list[str] | None) – The name of the velocity columns in the input data frame. These columns will be nested into the column velocity. If the list is empty or None, the nested velocity column will not be created. (default: None)
acceleration_columns (list[str] | None) – The name of the acceleration columns in the input data frame. These columns will be nested into the column acceleration. If the list is empty or None, the nested acceleration column will not be created. (default: None)
distance_column (str | None) – The name of the eye-to-screen distance column in the input data frame. If specified, the column will be used for pixel to dva transformations. If not specified, the constant eye-to-screen distance will be taken from the experiment definition. (default: None)
column_map (dict[str, str] | None) – The keys are the columns to read, the values are the names to which they should be renamed. (default: None)
add_columns (dict[str, str] | None) – Dictionary containing columns to add to loaded data frame. (default: None)
column_dtypes (dict[str, Any] | None) – Dictionary containing types for columns. (default: None)
**read_csv_kwargs (Any) – Additional keyword arguments to be passed to polars.read_csv() to read in the csv. These can include custom separators, a subset of columns, or specific data types for columns.

Returns:

The gaze data frame read from the csv file.

Return type:

GazeDataFrame

Notes

About using the arguments pixel_columns, position_columns, velocity_columns, and acceleration_columns:

By passing a list of columns as any of these arguments, these columns will be merged into a single column with the corresponding name , e.g. using pixel_columns will merge the respective columns into the column pixel.

The supported number of component columns with the expected order are:

zero columns: No nested component column will be created.
two columns: monocular data; expected order: x-component, y-component
four columns: binocular data; expected order: x-component left eye, y-component left eye, x-component right eye, y-component right eye,
six columns: binocular data with additional cyclopian data; expected order: x-component left eye, y-component left eye, x-component right eye, y-component right eye, x-component cyclopian eye, y-component cyclopian eye,

Examples

First let’s assume a CSV file stored tests/files/monocular_example.csv with the following content: shape: (10, 3) ┌──────┬────────────┬────────────┐ │ time ┆ x_left_pix ┆ y_left_pix │ │ — ┆ — ┆ — │ │ i64 ┆ i64 ┆ i64 │ ╞══════╪════════════╪════════════╡ │ 0 ┆ 0 ┆ 0 │ │ 1 ┆ 0 ┆ 0 │ │ 2 ┆ 0 ┆ 0 │ │ 3 ┆ 0 ┆ 0 │ │ … ┆ … ┆ … │ │ 6 ┆ 0 ┆ 0 │ │ 7 ┆ 0 ┆ 0 │ │ 8 ┆ 0 ┆ 0 │ │ 9 ┆ 0 ┆ 0 │ └──────┴────────────┴────────────┘

We can now load the data into a GazeDataFrame by specyfing the experimental setting and the names of the pixel position columns. We can specify a custom separator for the csv file by passing it as a keyword argument to polars.read_csv():

>>> from pymovements.gaze.io import from_csv
>>> gaze = from_csv(
...     file='tests/files/monocular_example.csv',
...     time_column = 'time',
...     time_unit='ms',
...     pixel_columns = ['x_left_pix','y_left_pix'],
...     separator = ',',
... )
>>> gaze.frame
shape: (10, 2)
┌──────┬───────────┐
│ time ┆ pixel     │
│ ---  ┆ ---       │
│ i64  ┆ list[i64] │
╞══════╪═══════════╡
│ 0    ┆ [0, 0]    │
│ 1    ┆ [0, 0]    │
│ 2    ┆ [0, 0]    │
│ 3    ┆ [0, 0]    │
│ …    ┆ …         │
│ 6    ┆ [0, 0]    │
│ 7    ┆ [0, 0]    │
│ 8    ┆ [0, 0]    │
│ 9    ┆ [0, 0]    │
└──────┴───────────┘

Please be aware that data types are inferred from a fixed number of rows. To ensure correct data types, you can pass a dictionary of column names and data types to the dtypes keyword argument of polars.read_csv():

>>> from pymovements.gaze.io import from_csv
>>> import polars as pl
>>> gaze = from_csv(
...     file='tests/files/monocular_example.csv',
...     time_column = 'time',
...     time_unit='ms',
...     pixel_columns = ['x_left_pix','y_left_pix'],
...     dtypes = {'time': pl.Int64, 'x_left_pix': pl.Int64, 'y_left_pix': pl.Int64},
... )
>>> gaze.frame
shape: (10, 2)
┌──────┬───────────┐
│ time ┆ pixel     │
│ ---  ┆ ---       │
│ i64  ┆ list[i64] │
╞══════╪═══════════╡
│ 0    ┆ [0, 0]    │
│ 1    ┆ [0, 0]    │
│ 2    ┆ [0, 0]    │
│ 3    ┆ [0, 0]    │
│ …    ┆ …         │
│ 6    ┆ [0, 0]    │
│ 7    ┆ [0, 0]    │
│ 8    ┆ [0, 0]    │
│ 9    ┆ [0, 0]    │
└──────┴───────────┘