Windshaft-cartodb-5.4.0/docs/aggregation.md

Tile Aggregation

To be able to represent a large amount of data (say, hundred of thousands to millions of points) in a tile. This can be useful both for raster tiles (where the aggregation reduces the number of features to be rendered) and vector tiles (the tile contais less features).

Aggregation is available only for point geometries. During aggregation the points are grouped using a grid; all the points laying in the same cell of the grid are summarized in a single aggregated result point.

• The position of the aggregated point is controlled by the placement parameter.
• The aggregated rows always contain at least a column, named _cdb_feature_count, which contains the number of the original points that the aggregated point represents.

Special default aggregation

When no placement or columns are specified a special default aggregation is performed.

This special mode performs only spatial aggregation (using a grid defined by the requested tile and the resolution, parameter, as all the other cases), and returns a random record from each group (grid cell) with all its columns and an additional _cdb_features_count with the number of features in the group.

Regarding the randomness of the sample: currently we use the row with the minimum cartodb_id value in each group.

The rationale behind having this special aggregation with all the original columns is to provide a mostly transparent way to handle large datasets without having to provide special map configurations for those cases (i.e. preserving the logic used to produce the maps with smaller datasets). Overviews have been used so far with this intent, but they are inflexible.

User defined aggregations

When either a explicit placement or columns are requested we no longer use the special, query; we use one determined by the placement (which will default to "centroid"), and it will have as columns only the aggregated columns specified, in addition to _cdb_features_count, which is always present.

We might decide in the future to allow sampling column values for any of the different placement modes.

Behaviour for raster and vector tiles

The vector tiles from a vector-only map will be aggregated by default. However, Raster tiles (or vector tiles from a map which defines CartoCSS styles) will be aggregated only upon request.

Aggregation that would otherwise occur can be disabled by passing an aggregation=false parameter to the map instantiation HTTP call.

To control how aggregation is performed, an aggregation option can be added to the layer:

{
    "layers": [
        {
            "options": {
                "sql": "SELECT * FROM data",
                "aggregation": {
                    "placement": "centroid",
                    "columns": {
                        "value": {
                            "aggregate_function": "sum",
                            "aggregated_column": "value"
                        }
                    }
                }
            }
        }
    ]
}

Even if aggregation is explicitly requested it may not be activated, e.g., if the geometries are not points or the whole dataset is too small. The map instantiation response contains metadata that informs if any particular layer will be aggregated when tiles are requested, both for vector (mvt) and raster (png) tiles.

{
  "layergroupid": "7b97b6e76590fef889b63edd2efb1c79:1513608333045",
  "metadata": {
    "layers": [
      {
        "type": "mapnik",
        "id": "layer0",
        "meta": {
          "stats": {
            "estimatedFeatureCount": 6232136
          },
          "aggregation": {
            "png": true,
            "mvt": true
          }
        }
      }
    ]
  }
}

Aggregation parameters

The aggregation parameters for a layer are defined inside an aggregation option of the layer:

{
    "layers": [
        {
            "options": {
                "sql": "SELECT * FROM data",
                "aggregation": {"...": "..."}
            }
        }
    ]
}

placement

Determines the kind of aggregated geometry generated:

point-sample

This is the default placement. It will place the aggregated point at a random sample of the grouped points, like the default aggregation does. No other attribute is sampled, though, the point will contain the aggregated attributes determined by the columns parameter.

point-grid

Generates points at the center of the aggregation grid cells (squares).

centroid

Generates points with the averaged coordinated of the grouped points (i.e. the points inside each grid cell).

columns

The aggregated attributes defined by columns are computed by a applying an aggregate function to all the points in each group. Valid aggregate functions are sum, avg (average), min (minimum), max (maximum) and mode (the most frequent value in the group). The values to be aggregated are defined by the aggregated column of the source data. The column keys define the name of the resulting column in the aggregated dataset.

For example here we define three aggregate attributes named total, max_price and price which are all computed with the same column, price, of the original dataset applying three different aggregate functions.

{
    "columns": {
        "total": { "aggregate_function": "sum", "aggregated_column": "price" },
        "max_price": { "aggregate_function": "max", "aggregated_column": "price" },
        "price": { "aggregate_function": "avg", "aggregated_column": "price" }
    }
}

Note that you can use the original column names as names of the result, but all the result column names must be unique. In particular, the names cartodb_id, the_geom, the_geom_webmercator and _cdb_feature_count cannot be used for aggregated columns, as they correspond to columns always present in the result.

resolution

Defines the cell-size of the spatial aggregation grid. This is equivalent to the CartoCSS -torque-resolution property of Torque maps.

The aggregation cells are resolution×resolution pixels in size, where pixels here are defined to be 1/256 of the (linear) size of a tile. The default value is 1, so that aggregation coincides with raster pixels. A value of 2 would make each cell to be 4 (2×2) pixels, and a value of 0.5 would yield 4 cells per pixel. In teneral values less than 1 produce sub-pixel precision.

Note that is independent of the number of pixels for raster tile or the coordinate resolution (mvt_extent) of vector tiles.

threshold

This is the minimum number of (estimated) rows in the dataset (query results) for aggregation to be applied. If the number of rows estimate is less than the threshold aggregation will be disabled for the layer; the instantiation response will reflect that and tiles will be generated without aggregation.

Example

{
    "version": "1.7.0",
    "extent": [-20037508.5, -20037508.5, 20037508.5, 20037508.5],
    "srid": 3857,
    "maxzoom": 18,
    "minzoom": 3,
    "layers": [
        {
            "type": "mapnik",
            "options": {
                "sql": "select * from table",
                "cartocss": "#table { marker-width: [total]; marker-fill: ramp(value, (red, green, blue), jenks); }",
                "cartocss_version": "2.3.0",
                "aggregation": {
                    "placement": "centroid",
                    "columns": {
                        "value": {
                            "aggregate_function": "avg",
                            "aggregated_column": "value"
                        },
                        "total": {
                            "aggregate_function": "sum",
                            "aggregated_column": "value"
                        }
                    },
                    "resolution": 2,
                    "threshold": 500000
                }
            }
        }
    ]
}