Specifying Data in Altair

Each top-level chart object (i.e. Chart, LayerChart, and VConcatChart, HConcatChart, RepeatChart, FacetChart) accepts a dataset as its first argument. The dataset can be specified in one of the following ways:

as a Pandas DataFrame
as a Data or related object (i.e. UrlData, InlineData, NamedData)
as a url string pointing to a json or csv formatted text file
as an object that supports the __geo_interface__ (eg. Geopandas GeoDataFrame, Shapely Geometries, GeoJSON Objects)

For example, here we specify data via a DataFrame:

                    import altair as alt
import pandas as pd

data = pd.DataFrame({'x': ['A', 'B', 'C', 'D', 'E'],
                     'y': [5, 3, 6, 7, 2]})
alt.Chart(data).mark_bar().encode(
    x='x',
    y='y',
)

                  

When data is specified as a DataFrame, the encoding is quite simple, as Altair uses the data type information provided by Pandas to automatically determine the data types required in the encoding.

By comparison, here we create the same chart using a Data object, with the data specified as a JSON-style list of records:

                    import altair as alt

data = alt.Data(values=[{'x': 'A', 'y': 5},
                        {'x': 'B', 'y': 3},
                        {'x': 'C', 'y': 6},
                        {'x': 'D', 'y': 7},
                        {'x': 'E', 'y': 2}])
alt.Chart(data).mark_bar().encode(
    x='x:O',  # specify ordinal data
    y='y:Q',  # specify quantitative data
)

                  

notice the extra markup required in the encoding; because Altair cannot infer the types within a Data object, we must specify them manually (here we use Encoding Shorthands to specify ordinal (O) for x and quantitative (Q) for y; see Encoding Data Types).

Similarly, we must also specify the data type when referencing data by URL:

                    import altair as alt
from vega_datasets import data
url = data.cars.url

alt.Chart(url).mark_point().encode(
    x='Horsepower:Q',
    y='Miles_per_Gallon:Q'
)

                  

We will further discuss encodings and associated types in Encodings, next.

Including Index Data

By design Altair only accesses dataframe columns, not dataframe indices. At times, relevant data appears in the index. For example:

                      import numpy as np
rand = np.random.RandomState(0)

data = pd.DataFrame({'value': rand.randn(100).cumsum()},
                    index=pd.date_range('2018', freq='D', periods=100))
data.head()

                    

                                         value
    2018-01-01  1.764052
    2018-01-02  2.164210
    2018-01-03  3.142948
    2018-01-04  5.383841
    2018-01-05  7.251399

                    

If you would like the index to be available to the chart, you can explicitly turn it into a column using the reset_index() method of Pandas dataframes:

                      alt.Chart(data.reset_index()).mark_line().encode(
    x='index:T',
    y='value:Q'
)

                    

If the index object does not have a name attribute set, the resulting column will be called "index". More information is available in the Pandas documentation.

Long-form vs. Wide-form Data

There are two common conventions for storing data in a dataframe, sometimes called long-form and wide-form. Both are sensible patterns for storing data in a tabular format; briefly, the difference is this:

wide-form data has one row per independent variable, with metadata recorded in the row and column labels.
long-form data has one row per observation, with metadata recorded within the table as values.

Altair’s grammar works best with long-form data, in which each row corresponds to a single observation along with its metadata.

A concrete example will help in making this distinction more clear. Consider a dataset consisting of stock prices of several companies over time. The wide-form version of the data might be arranged as follows:

                      wide_form = pd.DataFrame({'Date': ['2007-10-01', '2007-11-01', '2007-12-01'],
                          'AAPL': [189.95, 182.22, 198.08],
                          'AMZN': [89.15, 90.56, 92.64],
                          'GOOG': [707.00, 693.00, 691.48]})
print(wide_form)

                    

                                   Date    AAPL   AMZN    GOOG
2007-10-01  189.95  89.15  707.00
2007-11-01  182.22  90.56  693.00
2007-12-01  198.08  92.64  691.48

                    

Notice that each row corresponds to a single time-stamp (here time is the independent variable), while metadata for each observation (i.e. company name) is stored within the column labels.

The long-form version of the same data might look like this:

                      long_form = pd.DataFrame({'Date': ['2007-10-01', '2007-11-01', '2007-12-01',
                                   '2007-10-01', '2007-11-01', '2007-12-01',
                                   '2007-10-01', '2007-11-01', '2007-12-01'],
                          'company': ['AAPL', 'AAPL', 'AAPL',
                                      'AMZN', 'AMZN', 'AMZN',
                                      'GOOG', 'GOOG', 'GOOG'],
                          'price': [189.95, 182.22, 198.08,
                                     89.15,  90.56,  92.64,
                                    707.00, 693.00, 691.48]})
print(long_form)

                    

                                   Date company   price
2007-10-01    AAPL  189.95
2007-11-01    AAPL  182.22
2007-12-01    AAPL  198.08
2007-10-01    AMZN   89.15
2007-11-01    AMZN   90.56
2007-12-01    AMZN   92.64
2007-10-01    GOOG  707.00
2007-11-01    GOOG  693.00
2007-12-01    GOOG  691.48

                    

Notice here that each row contains a single observation (i.e. price), along with the metadata for this observation (the date and company name). Importantly, the column and index labels no longer contain any useful metadata.

As mentioned above, Altair works best with this long-form data, because relevant data and metadata are stored within the table itself, rather than within the labels of rows and columns:

                      alt.Chart(long_form).mark_line().encode(
  x='Date:T',
  y='price:Q',
  color='company:N'
)

                    

Wide-form data can be similarly visualized using e.g. layering (see Layered Charts), but it is far less convenient within Altair’s grammar.

If you would like to convert data from wide-form to long-form, there are two possible approaches: it can be done as a preprocessing step using pandas, or as a transform step within the chart itself. We will detail to two approaches below.

Converting Between Long-form and Wide-form: Pandas

This sort of data manipulation can be done as a preprocessing step using Pandas, and is discussed in detail in the Reshaping and Pivot Tables section of the Pandas documentation.

For converting wide-form data to the long-form data used by Altair, the melt method of dataframes can be used. The first argument to melt is the column or list of columns to treat as index variables; the remaining columns will be combined into an indicator variable and a value variable whose names can be optionally specified:

                        wide_form.melt('Date', var_name='company', value_name='price')

                      

                                     Date company   price
2007-10-01    AAPL  189.95
2007-11-01    AAPL  182.22
2007-12-01    AAPL  198.08
2007-10-01    AMZN   89.15
2007-11-01    AMZN   90.56
2007-12-01    AMZN   92.64
2007-10-01    GOOG  707.00
2007-11-01    GOOG  693.00
2007-12-01    GOOG  691.48

                      

For more information on the melt method, see the Pandas melt documentation.

In case you would like to undo this operation and convert from long-form back to wide-form, the pivot method of dataframes is useful.

                        long_form.pivot(index='Date', columns='company', values='price').reset_index()

                      

                            company        Date    AAPL   AMZN    GOOG
      2007-10-01  189.95  89.15  707.00
      2007-11-01  182.22  90.56  693.00
      2007-12-01  198.08  92.64  691.48

                      

For more information on the pivot method, see the Pandas pivot documentation.

Converting Between Long-form and Wide-form: Fold Transform

If you would like to avoid data preprocessing, you can reshape your data using Altair’s Fold Transform (see Fold Transform for a full discussion). With it, the above chart can be reproduced as follows:

                        alt.Chart(wide_form).transform_fold(
    ['AAPL', 'AMZN', 'GOOG'],
    as_=['company', 'price']
).mark_line().encode(
    x='Date:T',
    y='price:Q',
    color='company:N'
)

                      

Notice that unlike the pandas melt function we must explicitly specify the columns to be folded. The as_ argument is optional, with the default being ["key", "value"].

Generated Data

At times it is convenient to not use an external data source, but rather generate data for display within the chart specification itself. The benefit is that the chart specification can be made much smaller for generated data than for embedded data.

Sequence Generator

Here is an example of using the sequence() function to generate a sequence of x data, along with a Calculate Transform to compute y data.

                        import altair as alt

# Note that the following generator is functionally similar to
# data = pd.DataFrame({'x': np.arange(0, 10, 0.1)})
data = alt.sequence(0, 10, 0.1, as_='x')

alt.Chart(data).transform_calculate(
    y='sin(datum.x)'
).mark_line().encode(
    x='x:Q',
    y='y:Q',
)

                      

Graticule Generator

Another type of data that is convenient to generate in the chart itself is the latitude/longitude lines on a geographic visualization, known as a graticule. These can be created using Altair’s graticule() generator function. Here is a simple example:

                        import altair as alt

data = alt.graticule(step=[15, 15])

alt.Chart(data).mark_geoshape(stroke='black').project(
    'orthographic',
    rotate=[0, -45, 0]
)

                      

Sphere Generator

Finally when visualizing the globe a sphere can be used as a background layer within a map to represent the extent of the Earth. This sphere data can be created using Altair’s sphere() generator function. Here is an example:

                        import altair as alt

sphere_data = alt.sphere()
grat_data = alt.graticule(step=[15, 15])

background = alt.Chart(sphere_data).mark_geoshape(fill='aliceblue')
lines = alt.Chart(grat_data).mark_geoshape(stroke='lightgrey')

alt.layer(background, lines).project('naturalEarth1')

Geospatial Data

Working with geographical data in Altair is possible if the object contains a __geo_interface__ attribute. This attribute represents the geo_interface which is a Python protocol for Geospatial Data. The protocol follows a GeoJSON-like structure to store geo-spatial vector data.

To make working with Geospatial Data as similar as working with long-form structured data the geo_interface is serialized in order to: - make it be correctly interpreted by Altair - provide users a similar experience as when working with tabular data such as Pandas.

Altair can interpret a spatial bounded entity (a Feature) or a list of Features (FeatureCollection). In order for correct interpretation it is made sure that all records contain a single geometry (one of Point, LineString, Polygon, MultiPoint, MultiLineString, MultiPolygon, and GeometryCollection) and is stored as a Feature entity.

The most basic Feature is an entity that only contains a Geometry object. For example a Polygon:

                      {
    "type": "Feature",
    "geometry": {
        "coordinates": [[
            [0, 0],
            [0, 2],
            [2, 2],
            [2, 0],
            [0, 0]
        ]],
        "type": "Polygon"
    }
}

                    

Often, the Feature contains also additional metadata next to the Geometry object. The __geo_interface__ provides two approaches to store metadata. - Metadata stored as a dictionary within the key properties (so called properties member). This properties member must exist in a valid Feature entity. - Metada may be stored directly as foreign members on the top-level of the Feature. There is no normative processing model for usage of this declaration.

Altair serializes the metadata from the properties in combination with the declared geometry as Feature entities. The result of this approach is that the keys type and geometry in the properties member will be overwritten if used.

So a __geo_interface__ that is registered as such

                      {
    "type": "Feature",
    "id": "f1",
    "geometry": {...},
    "properties": {
        "id": 1,
        "foo": "xx",
        "bah": "yy",
        "type": "zz"
    },
    "title": "Example Feature"
}

                    

Is serialized as such:

                      {
    "type": "Feature",
    "geometry": {...},
    "foo": "xx",
    "bah": "yy",
    "id": 1
}

                    

The nested “type”: “zz” in the properties member is overwritten by “type”:”Feature” and only the metadata stored in the properties member is serialized. Meaning that foreign members and the commonly used identifier are not serialized.

GeoPandas vs Pandas

A GeoDataFrame is a DataFrame including a special column with spatial geometries. The column-name containing the spatial geometries defaults to geometry. To directly use a GeoDataFrame with Altair means in practice that only the column-name type should be avoided.

Projections

Altair works best when the Geospatial Data adopts the World Geodetic System 1984 as its geographic coordinate reference system with units in decimal degrees.

Try to avoid putting projected data into Altair, but reproject your spatial data to EPSG:4326 first.

If your data comes in a different projection (eg. with units in meters) and you don’t have the option to reproject the data, try using the project configuration (type: ‘identity’, reflectY’: True). It draws the geometries in a cartesian grid without applying a projection.

Winding order

LineString, Polygon and MultiPolygon geometries contain coordinates in an order: lines go in a certain direction, and polygon rings do too. The GeoJSON-like structure of the __geo_interface__ recommends the right-hand rule winding order for Polygon and MultiPolygons. Meaning that the exterior rings should be counterclockwise and interior rings are clockwise. While it recommends the right-hand rule winding order, it does not reject geometries that do not use the right-hand rule.

Altair does NOT follow the right-hand rule for geometries, but uses the left-hand rule. Meaning that exterior rings should be clockwise and interior rings should be counterclockwise.

If you face a problem regarding winding order, try to force the left-hand rule on your data before usage in Altair using GeoPandas for example as such:

                      from shapely.ops import orient # version >=1.7a2
gdf.geometry = gdf.geometry.apply(orient, args=(-1,))