Analytics Toolbox for Snowflake

Analytics Toolbox for Snowflake

Github Code

tiler

We currently provide procedures to create two types of tilesets: simple and aggregation tilesets, the former to visualize features individually and the latter to generate aggregated point visualizations. Visit the Overview section to learn more about tileset types.

CREATE_POINT_AGGREGATION_TILESET

Description

Generates a point aggregation tileset.

  • input: VARCHAR that can either contain a table name (e.g. database.schema.tablename) or a full query (e.g.(SELECT * FROM database.schema.tablename)).
  • output_table: VARCHAR of the format database.schema.tablename where the resulting tileset will be stored. The database and schema must exist and the caller needs to have permissions to create a new table in it. The process will fail if the table already exists.
  • options: VARCHAR containing a valid JSON with the different options. Valid options are described in the table below.
Option Description
geom_column Default: "GEOM". A VARCHAR that indicates the name of the geography column that will be used. The geography column must be of type GEOGRAPHY and contain only points. The capitalization (uppercase/lowercase letters) of the name must match exactly the column name; note that Snowflake by default converts names to uppercase.
zoom_min Default: 0. An INTEGER that defines the minimum zoom level at which tiles will be generated. Any zoom level under this level won’t be generated.
zoom_max Default: 15; maximum: 20. An INTEGER that defines the maximum zoom level at which tiles will be generated. Any zoom level over this level won’t be generated.
aggregation_resolution Default: 6. An INTEGER that specifies the resolution of the spatial aggregation.

Aggregation for zoom z is based on quadgrid cells at z + resolution level. For example, with resolution 6, the z0 tile will be divided into cells that match the z6 tiles, or the cells contained in the z10 tile will be the boundaries of the z16 tiles within them. In other words, each tile is subdivided into 4^resolution cells, which is the maximum number of resulting features (aggregated) that the tiles will contain.

Note that adding more granularity necessarily means heavier tiles which take longer to be transmitted and processed in the final client, and you are more likely to hit the internal memory limits.
aggregation_placement Default: "cell-centroid". A VARCHAR that defines what type of geometry will be used to represent the cells generated in the aggregation, which will be the features of the resulting tileset. There are currently four options:
  • "cell-centroid": Each feature will be defined as the centroid of the cell, that is, all points that are aggregated together into the cell will be represented in the tile by a single point positioned at the centroid of the cell.
  • "cell": Each feature will be defined as the entire cell’s polygon, thus the final representation in the tile will be a polygon. This provides more precise coordinates but takes more space in the tile and requires more CPU to process it in the renderer.
  • "features-any": The aggregation cell will be represented by any random point from the source data contained within it. That is, if 10 points fall inside a cell, the procedure will randomly choose the location of one of them to represent the aggregation cell.
  • "features-centroid": The feature will be defined as the centroid (point) of the collection of points within the cell.
metadata Default: {}. A JSON object to specify the associated metadata of the tileset. Use this to set the name, description and legend to be included in the TileJSON. Other fields will be included in the object extra_metadata.
properties Default: {}. A JSON object that defines the properties that will be included associated with each cell feature. Each property is defined by its name, type (Number, Boolean, String, etc.) and formula to be applied to the values of the points that fall under the cell. This formula can be any SQL formula that uses an aggregate function supported by Snowflake and returns the expected type. Note that every property different from Number will be casted to String.
max_tile_features Default: 10000. A NUMBER that sets the maximum number of features (points) a tile can contain. When this maximum is reached, the procedure will drop features according to the chosen max_tile_size_strategy. You can configure in which order the features are kept by setting the tile_feature_order property. Any value lower than 4^aggregation_resolution will be ineffective, therefore the default of 10000 only applies if aggregation_resolution is higher than 6.
tile_feature_order Default: RANDOM(). A STRING defining the order in which features are added to a tile. This expects the SQL ORDER BY keyword definition, such as "aggregated_total DESC". The "ORDER BY" part must not be included. You can use any source column even if it is not included in the tileset as a property.
max_tile_size_strategy Default: "throw_error". A STRING that specifies how to apply the limit defined by max_tile_features. There are four options available:
  • "drop_features": In each tile the features that would exceed the limit are dropped. Different fractions of the total features may be dropped in each tile, which on a map can appear as noticeable differences in feature density between tiles.
  • "drop_fraction_as_needed": For every zoom level, this process will drop a consistent fraction of features in every tile to make sure all generated tiles are below the limit. Since a constant fraction of the features is dropped for all tiles of a given zoom level, this will in general drop more features in less populated tiles than the the "drop_features" strategy.
  • "return_null": A row with a NULL DATA column will be produced for all tiles that exceed the limit.
  • "throw_error": The procedure execution will be aborted if any tile exceeds the limit.

. For the drop_ strategies, features will be retained according to the tile_feature_order specified.

Result

The generated tileset consists of a table with the following columns, where each row represents a tile:

  • Z: zoom level of the tile.
  • X: X-index of the tile (0 to 2^Z-1).
  • Y: Y-index of the tile (0 to 2^Z-1).
  • DATA: contents of the tile, encoded as a GeoJSON string (a feature collection). It will contain the resulting points (location of the aggregated features) and their attributes (as defined by properties).

Additionally, there is a row identified with Z=-1 which contains metadata about the tileset in the DATA column in JSON format. It contains the following properties:

  • bounds: geographical extents of the source as a string in Xmin, Ymin, Xmax, Ymax format.
  • center: center of the geographical extents as X, Y, Z, where the Z represents the zoom level where a single tile spans the whole extents size.
  • zmin: minimum zoom level in the tileset.
  • zmax: maximum zoom level in the tileset.
  • tilestats: stats about the feature’s properties. In addition to its name (attribute) and type, it contains min, max, average, sum and quantiles for numeric attributes and categories for text attributes.

Example

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

CALL carto.CREATE_POINT_AGGREGATION_TILESET(
  'SELECT * FROM DATABASE.SCHEMA.CITIES_TABLE',
  'DATABASE.SCHEMA.CITIES_TILESET',
  '{
    "geom_column": "GEOM",
    "zoom_min": 0,
    "zoom_max": 12,
    "aggregation_resolution": 5,
    "aggregation_placement": "cell-centroid",
    "properties": {
      "NUM_CITIES": {
        "formula": "COUNT(*)",
        "type": "Number"
      },
      "POPULATION_SUM": {
        "formula": "SUM(POPULATION)",
        "type": "Number"
      },
      "CITY_NAME": {
        "formula": "IFF(COUNT(*) <= 1, ANY_VALUE(CITY_NAME), NULL)",
        "type": "String"
      }
    },
    "metadata": {
      "name": "Population",
      "description": "Population in the cities"
    }
  }'
)

In the example above, for all features we would get a property "NUM_CITIES" with the number of points that fall in it and "POPULATION_SUM" with the sum of the population in those cities. In addition to this, when there is only one point that belongs to this property (and only in that case) we will also get the column values from the source data in "CITY_NAME".

CREATE_SIMPLE_TILESET

Description

Generates a simple tileset.

  • input: VARCHAR that can either contain a table name (e.g. database.schema.tablename) or a full query (e.g.‘SELECT * FROM db.schema.tablename’).
  • output_table: VARCHAR of the format ‘database.schema.tablename’ where the resulting tileset will be stored.
  • options: VARCHAR containing a valid JSON with the different options. Valid options are described in the table below.
Option Description
geom_column Default: "GEOM". A VARCHAR that specifies the name of the geography column that will be used. It must be of type GEOGRAPHY. The capitalization (uppercase/lowercase letters) of the name must match exactly the column name; note that Snowflake by default will use only uppercase letters for the column names, but this can be altered if column names are quoted in their definition. Do not use quotes here, just mutch the capitalization.
zoom_min Default: 0. A NUMBER that defines the minimum zoom level at which tiles will be generated. Any zoom level under this level won’t be generated.
zoom_max Default: 10. A NUMBER that defines the maximum zoom level at which tiles will be generated. Any zoom level over this level won’t be generated.
metadata Default: {}. A JSON object to specify the associated metadata of the tileset. Use this to set the name, description and legend to be included in the TileJSON. Other fields will be included in the object extra_metadata.
properties Default: {}. A JSON object that defines the properties that will be included associated with each cell feature. Each property is defined by its name and type (Number, String, etc.). Please note that every property different from Number will be casted to String. Property names must correspond to column names in the input and match it’s capitalization exactly; note by default Snowflake will use only uppercase letters for the column names but this can altered if column names where quoted in their definitions. Do not use quotes here, just mutch the capitalization. The properties will appear in the GeoJSON data with the same capitalization as the column names.
max_tile_features Default: 10000. A NUMBER that sets the maximum number of features a tile can contain. This limit only applies when the input geometries are points. When this limit is reached, the procedure will stop adding features into the tile. You can configure in which order the features are kept by setting the tile_feature_order property.
max_tile_vertices Default: 200000. A NUMBER that sets the maximum number of vertices a tile can contain. This limit only applies when the input geometries are lines or polygons. When this maximum is reached, the procedure will drop features according to the chosen max_tile_size_strategy. You can configure in which order the features are kept by setting the tile_feature_order property.
tile_feature_order Default: RANDOM() for points, ST_AREA() DESC for polygons, ST_LENGTH() DESC for lines. A STRING defining the order in which properties are added to a tile. This expects the SQL ORDER BY keyword definition, such as "aggregated_total DESC". The "ORDER BY" part must not be included. You can use any source column even if it is not included in the tileset as a property.
max_tile_size_strategy Default: "throw_error". A STRING that specifies how to apply the limit defined by max_tile_features or max_tile_vertices. There are four options available:
  • "drop_features": In each tile the features that exceed the limit are dropped. Different fractions of the total features may be dropped in each tile, which on a map can appear as noticeable differences in feature density between tiles.
  • "drop_fraction_as_needed": For every zoom level, this process will drop a consistent fraction of features in every tile to make sure all generated tiles are below the limit. Since a constant fraction of the features is dropped for all tiles of a given zoom level, this will in general drop more features in less populated tiles than the the "drop_features" strategy.
  • "return_null": A row with a NULL data column will be produced for all tiles that exceed the limit.
  • "throw_error": The procedure execution will be aborted if any tile exceeds the limit.

. For the drop_ strategies, features will be retained according to the tile_feature_order specified.
generate_feature_id Default: true. A BOOLEAN used to add a unique numeric id in the GeoJSON.

Example

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12

CALL carto.CREATE_SIMPLE_TILESET(
  'SELECT geom, population, category FROM mypopulationtable',
  'MYDB.MYSCHEMA.population_tileset',
  '{
    "geom_column": "GEOM",
    "zoom_min": 0, "zoom_max": 6,
    "properties": {
      "POPULATION": "Number",
      "category": "String"
    }
  }'
)

CREATE_SPATIAL_INDEX_TILESET

Description

Creates a tileset that uses a spatial index (H3 and QUADBIN are currently supported), aggregating data from an input table that uses that same spatial index.

Aggregated data is computed for all levels between resolution_min and resolution_max. For each resolution level, all tiles for the area covered by the source table are added, with data aggregated at level resolution + aggregation resolution.

  • source_table: STRING that can either be a quoted qualified table name (e.g. `database.schema.tablename`) or a full query contained by parentheses (e.g.(SELECT * FROM `database.schema.tablename`)).
  • target_table: Where the resulting table will be stored. It must be a STRING of the form `projectID.dataset.tablename`. The projectID can be omitted (in which case the default one will be used). The dataset must exist and the caller needs to have permissions to create a new table on it. The process will fail if the target table already exists.
  • options: STRING containing a valid JSON with the different options. Valid options are described the table below.
    Option Description
    resolution_min Default: 2. A NUMBER that defines the minimum resolution level for tiles. Any resolution level under this level won’t be generated.
    resolution_max Default: 15. A NUMBER that defines the maximum resolution level for tiles. Any resolution level over this level won’t be generated.
    spatial_index_column A STRING in the format spatial_index_type:column_name, with spatial_index_type being the type of spatial index used in the input table (can be quadbin or h3), and column_name being the name of the column in that input table that contains the tile ids. Notice that the spatial index name is case-sensitive. The type of spatial index also defines the type used in the output table, which will be QUADBIN (for spatial index type quadbin) or H3 (for spatial index type h3).
    resolution A NUMBER defining the resolution of the tiles in the input table.
    aggregation_resolution Defaults: 6 for QUADBIN tilesets, 4 for H3 tilesets. A NUMBER defining the resolution to use when aggregating data at each resolution level. For a given resolution, data is aggregated at resolution_level + aggregation resolution.
    properties A JSON object containing the aggregated properties to add to each tile in the output table. It cannot be empty, since at least one property is needed for aggregating the original values. Properties are case sensitive.
    metadata Default: {}. A JSON object to specify the associated metadata of the tileset. Use this to set the name, description and legend to be included in the TileJSON. Other fields will be included in the object extra_metadata.

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

CALL carto.CREATE_SPATIAL_INDEX_TILESET(
  'YOUR_DATABASE.YOUR_SCHEMA.INPUT_TABLE_QUADBIN_LEVEL14',
  'YOUR_DATABASE.YOUR_SCHEMA.OUTPUT_TILESET_QUADBIN_LEVEL14',
  '{
    "spatial_index_column": "quadbin:INDEX",
    "resolution": 14,
    "resolution_min": 0,
    "resolution_max": 8,
    "aggregation_resolution": 6,
    "properties": {
      "POPULATION": {
        "formula": "SUM(POPULATION)",
        "type": "Number"
      }
    }
  }'
);

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17

CALL carto.CREATE_SPATIAL_INDEX_TILESET(
  '(SELECT * FROM YOUR_DATABASE.YOUR_SCHEMA.INPUT_TABLE_H3_LEVEL10)',
  'YOUR_DATABASE.YOUR_SCHEMA.OUTPUT_TILESET_H3_LEVEL10',
  '{
    "spatial_index_column": "h3:INDEX",
    "resolution": 10,
    "resolution_min": 0,
    "resolution_max": 6,
    "aggregation_resolution": 4,
    "properties": {
      "POPULATION": {
        "formula": "SUM(POPULATION)",
        "type": "Number"
      }
    }
  }'
);

Snowflake treats columns uppercase by default, to set explicit lowercase use the following syntax:

  {
    "spatial_index_column": "quadbin:\\"index\\"",
    "properties": {
      "population": {
        "formula": "SUM(\\"population\\")",
        "type": "Number"
      }
    }
  }