`. The 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.
{% hint style="warning" %}
**warning**
If a query is passed in `input`, it might be evaluated multiple times to generate the tileset. Thus, non-deterministic functions, such as \[`ROW_NUMBER`] should be avoided. If such a function is needed, the query should be saved into a table first and then passed as `input`, to avoid inconsistent results.
{% endhint %}
| Option | Description |
| ------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `geom_column` | Default: `"geom"`. A `VARCHAR` that indicates the name of the geometry column that will be used. This column must be of type `GEOMETRY` with an SRID 4326 and contain only points. |
| `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: `12`; 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 + 1 = 7 @ tile\_resolution of 1 (default). 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](https://github.com/mapbox/tilejson-spec/tree/master/2.2.0). 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](https://docs.aws.amazon.com/redshift/latest/dg/c_Aggregate_Functions.html) supported by Redshift and returns the expected type. Note that every property different from Number will be casted to String. |
| `max_tile_features` | Default: `10000 * 4 = 40000` @ `tile_resolution` of `1` (default). 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 8. |
| `tile_feature_order` | Default: `""` (disabled). A `VARCHAR` 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. Please note that the default behavior will add features to the tile according to the order in which they appear in the input table or query. |
| `max_tile_size_strategy` | Default: "throw\_error". A VARCHAR that specifies how to apply the limit defined by max\_tile\_features. There are two 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."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.
|
| `max_categories_limit` | Default: `10`. A `NUMBER` that sets the maximum number of categories a property can have. |
| `tile_resolution` | Default: `1`. A `FLOAT` which determines final tile resolution. Valid values are 0.25, 0.5, 1, 2 and 4 which correspond to tile sizes of 256px, 512px, 1024px, 2048px and 4096px respectively. |
In web map tilesets, each additional zoom level has 4 times the amount of tiles as the previous zoom. Level 0 has 1 tile, level 1 has 4, level 2 has 16, etc.
For the map viewer, `tile_resolution` is really a way of using an offset zoom level to load 'large' but few tiles , or 'small' but many tiles. For example, at zoom level 2 we might have to load 16 tiles to fill our screen. By increasing `tile_resolution` one step (eg. `0.5` to `1`), we artificially use one z-level less (zoom level of 1) to load 4 (larger) tiles. Or we could increase `tile_resolution` two steps (eg. from 0.5 to 2), to artificially use two z-levels less (zoom level of 0) and thus load just one (even larger) tile. Here is a table which illustrates the *real* requested Z levels based on a tilesets `tile_resolution`.
| Map Zoom | TR 0.25 (256px) | TR 0.5 (512px) | TR 1 (1024px) | TR 2 (2048px) | TR 4 (4096px) |
| :------: | :-------------: | :------------: | :-----------: | :-----------: | :-----------: |
| 1 | 2 | 1 | 0 | 0 | 0 |
| 2 | 3 | 2 | 1 | 0 | 0 |
| 3 | 4 | 3 | 2 | 1 | 0 |
| 4 | 5 | 4 | 3 | 2 | 1 |
As shown, `tile_resolution` of `0.5` is where the tileset zoom level and map zoom levels match, so we use `0.5` as our baseline even though the default `tile_resolution` is `1`. Other `tile_resolution` values (eg, 1, 2, 4) will use offset z-levels when loaded on the map.
## Relationship between `tile\_resolution` and `aggregation\_resolution`
The value of `aggregation_resolution` will be adjusted based on `tile_resolution`. Although the default `tile_resolution` is 1, we use `0.5` as the baseline. So by default, `aggregation_resolution` gets adjusted. Its value (whether default or user-specified) will be decreased/increased as outlined below:
| tile\_resolution | Adjustment | Default | User-supplied Eg. 8 |
| :--------------: | :--------: | :-------: | :-----------------: |
| 0.25 | -1 | 6 - 1 = 5 | 8 - 1 = 7 |
| 0.5 | 0 | 6 + 0 = 6 | 8 + 0 = 8 |
| 1 | +1 | 6 + 1 = 7 | 8 + 1 = 9 |
| 2 | +2 | 6 + 2 = 8 | 8 + 2 = 10 |
| 4 | +3 | 6 + 3 = 9 | 8 + 3 = 11 |
Such that, 7, the default `aggregation_level` @ `tile_resolution` of `1`, is better thought of as 6 + 1 = 7. Likewise, if a user specified an `aggregation_resolution` of 8, and `tile_resolution` of 4, the generated tile will actually use 8 + 3 = 11. But, when rendered on the map, the z-levels are offset by -3 for `tile_resolution` 4 so the tiles are geographically larger but otherwise look the same as those generated with another `tile_resolution` value.
## Relationship between `tile\_resolution` and `max\_tile\_features`
At the default `tile_resolution` of `1`, any values set for option(s) `max_tile_features` will be multiplied by 4. Even when unspecified, the default values for `max_tile_features` (40000) is better thought of as the default of `10000` @ `tile_resolution` of `0.5`, multiplied by 4. Such that, `max_tile_features` @ `tile_resolution` of `1` => \[default max\_tile\_features @ 0.5] X 4 => 10000 X 4 = 40000.
This factor increases to 16 for `tile_resolution` of `2` and 64 for `tile_resolution` of `4`. Likewise, it decreases to 1 (ie. no change) at `tile_resolution` 0.5, and 0.25 at `tile_resolution` of `0.25` (ie. divide by 4).
Although this is somewhat unintuitive, the offset ensures that tilesets generated using the same options (but different tile\_resolutions) will always appear the same on the map.
{% hint style="warning" %}
**FEATURES PER TILE LIMITS**
The value of `aggregation_resolution` sets an upper bound to how many features can be present in a tile. For a value of *n*, a maximum of 4^*n* (4 raised to n) features can be present in a tile. For example, for an aggregation resolution of 8, the maximum number of features (points) will be 65536 per tile. This value can be too high and produce tiles that are too large when either the aggregation resolution is high or many properties are included. In that case, to improve the performance of the map visualizations, the `max_tile_features` should be used to limit the size of the tiles to about 1MB.
{% endhint %}
**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 in the `data` column identified by `Z=-1` which contains metadata about the tileset 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` and `sum`.
**Example**
{% code overflow="wrap" lineNumbers="true" %}
```sql
CALL carto.CREATE_POINT_AGGREGATION_TILESET(
'SELECT * FROM .',
'.',
'{
"geom_column": "geom",
"zoom_min": 0,
"zoom_max": 12,
"aggregation_resolution": 5,
"aggregation_placement": "cell-centroid",
"tile_resolution": 0.5,
"properties": {
"num_cities": {
"formula": "COUNT(*)",
"type": "Number"
},
"population_sum": {
"formula": "SUM(population)",
"type": "Number"
},
"city_name": {
"formula": "(CASE WHEN COUNT(*) <= 1 THEN ANY_VALUE(city_name) ELSE NULL END)",
"type": "String"
}
},
"metadata": {
"name": "Population",
"description": "Population in the cities"
}
}'
);
```
{% endcode %}
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"`.
**Example**
{% code overflow="wrap" lineNumbers="true" %}
```sql
CALL CREATE_SIMPLE_TILESET(
'SELECT geom, do_label, do_area FROM usa_polygons_simplified',
CURRENT_DATABASE() || '.' || CURRENT_SCHEMA() || '.' || 'tileset__usa_polygons__res4',
'{
"geom_column": "geom",
"zoom_min": 0,
"zoom_max": 12, "max_tile_size_strategy": "drop_features", "tile_resolution": 4,
"properties": {
"do_label": "String", "do_area": "Number"
}
}'
);
```
{% endcode %}
In the above example a simple tileset is created from a polygon table. The tile\_resolution of 4 will means the tiles will be 4096px when rendered.
## CREATE\_SPATIAL\_INDEX\_TILESET
```sql
CREATE_SPATIAL_INDEX_TILESET(input, output_table, options)
```
**Description**
Creates a tileset that uses a spatial index (H3 and QUADKEYS 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`.
**Input parameters**
* `input`: `VARCHAR` that can either contain a table name (e.g. `.`) or a full query (e.g.`'SELECT * FROM .'`).
* `output_table`: `VARCHAR` name of the output table, e.g. `.`. The 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.
{% hint style="warning" %}
**warning**
If a query is passed in `input`, it might be evaluated multiple times to generate the tileset. Thus, non-deterministic functions, such as \[`ROW_NUMBER`] should be avoided. If such a function is needed, the query should be saved into a table first and then passed as `input`, to avoid inconsistent results.
{% endhint %}
| Option | Description |
| ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `resolution_min` | Default: `0`. A `NUMBER` that defines the minimum resolution level for tiles. Any resolution level under this level won't be generated. |
| `resolution_max` | Default: `12` for QUADBIN tilesets, `6` for H3 tilesets. A `NUMBER` that defines the maximum resolution level for tiles. Any resolution level over this level won't be generated. |
| `spatial_index_column` | A `VARCHAR` 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 QUADKEY 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`. |
| `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](https://github.com/mapbox/tilejson-spec/tree/master/2.2.0). Other fields will be included in the object extra\_metadata. |
| `max_categories_limit` | Default: `10`. A `NUMBER` that sets the maximum number of categories a property can have. |
| `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, Boolean, String, etc.). Please note that every property different from Number will be casted to String. |
| `tile_resolution` | Default: `1`. Valid values are `0.25` (256px), `0.5` (512px), `1` (1024px), `2` (2048px) or `4` (4096px) |
**Examples**
{% code overflow="wrap" lineNumbers="true" %}
```sql
CALL carto.CREATE_SPATIAL_INDEX_TILESET(
'.',
'.',
'{
"spatial_index_column": "quadbin:index",
"resolution": 14,
"resolution_min": 0,
"resolution_max": 8,
"aggregation_resolution": 6,
"properties": {
"population": {
"formula": "SUM(population)",
"type": "Number"
}
}
}'
);
```
{% endcode %}
{% code overflow="wrap" lineNumbers="true" %}
```sql
CALL carto.CREATE_SPATIAL_INDEX_TILESET(
'SELECT * FROM .',
'.',
'{
"spatial_index_column": "h3:index",
"resolution": 10,
"resolution_min": 0,
"resolution_max": 6,
"aggregation_resolution": 4,
"properties": {
"population": {
"formula": "SUM(population)",
"type": "Number"
}
}
}'
);
```
{% endcode %}
{% hint style="info" %}
**Additional examples**
* [Creating spatial index tilesets](https://academy.carto.com/advanced-spatial-analytics/spatial-analytics-for-redshift/step-by-step-tutorials/creating-spatial-index-tilesets)
{% endhint %}