@ -8,7 +8,7 @@ To build a query with autocomplete, you need a `text` parameter, representing wh
There are two user experience pitfalls to watch out for when implementing a client-side typeahead solution:
**Requests must be throttled.** Since autocomplete requests generally correspond directly to user input, it's important to account for fast typers and throttle requests when using the autocomplete endpoint. Mapzen Search has a per second rate limit (that defaults to 6 requests per second), but some devices and networks (for example, mobile phones on a slow connection) may also respond poorly when too many requests are sent too quickly, so be sure to do some testing on your own. [Learn more in this interactive demo.](http://jsfiddle.net/missinglink/19e2r2we/)
**Requests must be throttled.** Since autocomplete requests generally correspond directly to user input, it's important to account for fast typers and throttle requests when using the autocomplete endpoint. Some devices and networks (for example, mobile phones on a slow connection) may respond poorly when too many requests are sent too quickly, so be sure to do some testing on your own. [Learn more in this interactive demo.](http://jsfiddle.net/missinglink/19e2r2we/)
**Responses are asynchronous.** This means you cannot be sure responses will be returned in the same order they were requested. If you were to send two queries synchronously, first `'Lo'` then `'London'`, you may find the `'London'` response would arrive before the `'Lo'` response. This will result in a quick flash of `'London'` results followed by the results for `'Lo'`, which can confuse the user.
@ -19,7 +19,7 @@ To focus your search based upon a geographical area, such as the center of the u
The `/autocomplete` endpoint can promote nearby results to the top of the list, while still allowing important matches from farther away to be visible. For example, searching `hard rock cafe` with a focus on Berlin:
> [/v1/autocomplete?api_key=mapzen-xxxxxx&__focus.point.lat=52.5&focus.point.lon=13.3&text=hard rock cafe__](https://search.mapzen.com/v1/autocomplete?focus.point.lat=52.5&focus.point.lon=13.3&text=hard rock cafe)
> [/v1/autocomplete?api_key=your-mapzen-api-key&__focus.point.lat=52.5&focus.point.lon=13.3&text=hard rock cafe__](https://mapzen.github.io/search-sandbox/?query=autocomplete&focus.point.lat=52.5&focus.point.lon=13.3&text=hard rock cafe)
with `focus.point` you will find the Berlin restaurant first:
```
@ -65,7 +65,7 @@ The `sources` parameter allows you to specify from which data sources you'd like
Note that you need an actual `gid` value to make a `/place` search. For example, if you search for an address and the result is [interpolated](addresses.md#address-interpolation), then there is no discrete `gid` to use for a `/place` search because interpolated results may be from multiple data sources.
@ -14,14 +14,14 @@ Note that you need an actual `gid` value to make a `/place` search. For example,
To search for more than one `/place` in a request, join multiple values together and separate them with a comma. For example, this `/place` query looks up the Eiffel Tower in OpenStreetMap and the borough of Manhattan in Who's on First:
If you enter a valid `gid` that cannot be found or has "expired" due to a newer build, you may get empty results. The request will NOT return an error.
If the structure of your `gid` is invalid, an error will be returned as part of the geojson structure.
If the structure of your `gid` is invalid, an error will be returned as part of the GeoJSON structure.
Keep in mind that if you enter a `gid` that cannot be found in a list of multiple ids, then the `features` array in the response contains a different number of elements than the number of requests. For example, your request may have three IDs requested but only two results returned. The reason for this is that the `features` section of the response is GeoJSON-compliant, and JSON does not allow a way to convey an exception condition (not even an empty JSON element, `{}`). For this reason, if your application is dependent upon the results mapping directly to the individual input requests in order, then you'll have to do your own bookkeeping to handle exception conditions.
Keep in mind that if you enter a `gid` that cannot be found in a list of multiple IDs, then the `features` array in the response contains a different number of elements than the number of requests. For example, your request may have three IDs requested but only two results returned. The reason for this is that the `features` section of the response is GeoJSON-compliant, and JSON does not allow a way to convey an exception condition (not even an empty JSON element, `{}`). For this reason, if your application is dependent upon the results mapping directly to the individual input requests in order, then you'll have to do your own bookkeeping to handle exception conditions.
:warning: It is important to not use any `gid` values to attempt `/place` queries after longer than an hour. These ids are not intended to be stable across build, as we employ datasets that do not have consistent ids.
:warning: It is important to not use any `gid` values to attempt `/place` queries after longer than an hour. These IDs are not intended to be stable across builds, as datasets are used that do not have consistent IDs.
@ -68,9 +68,10 @@ Additionally, [/reverse](reverse.md) queries will have a `distance` parameter, w
All results returned from Mapzen Search are points, and can be found in the `coordinates` array. Following the [GeoJSON specification](http://geojson.org/geojson-spec.html#positions), these coordinates are in **longitude, latitude** order.
### `gid`
All places in Mapzen Search have a global identifier, known as a `gid`. Each matching record returned from a [/search](search.md), [/autocomplete](autocomplete.md), or [/reverse](reverse.md) geocoding request has a `gid` field.
The `gid` consists of a `layer` (such as `address` or `country`), an identifier for the original data source (such as `openstreetmap` or `openaddresses`), and an `id` for the individual record corresponding to the original source idenfier, where possible. This information is also available as properties on the individual results as `layer`, `source`, and `source_id`.
The `gid` consists of a `layer` (such as `address` or `country`), an identifier for the original data source (such as `openstreetmap` or `openaddresses`), and an `id` for the individual record corresponding to the original source identifier, where possible. This information is also available as properties on the individual results as `layer`, `source`, and `source_id`.
#### :warning: Follow these guidelines regarding the `gid`:
@ -79,9 +80,11 @@ The `gid` consists of a `layer` (such as `address` or `country`), an identifier
- You should not attempt to parse `gid` strings for information or store them for future use. You should only use `gid` at the time when you receive the search results. One valid use for the `gid` is to retrieve full details on a particular result from the [/place](place.md) endpoint.
### `label`
The `label` is a human-friendly representation of the place, ready to be displayed to an end user. The label field attempts to use a format that is right for the region the result is in, although Mapzen Search only supports a few countries at the moment.
### `confidence`
The confidence score is an estimation of how accurately this result matches the query.
For the [/reverse](reverse.md) endpoint, the confidence score is determined solely by its distance from the coordinate specified. Closer results get a higher score.
@ -91,6 +94,7 @@ For the [/search](search.md) endpoint, it primarily takes into account how well
Additionally, the confidence score can optionally be biased along with other results, like test scores in a classroom might be graded on a curve. This takes into account both the property matches described above and the distance between results. This relative scoring is enabled on Mapzen Search, but can be turned off when hosting your own Pelias instance.
### `bbox`
Features from Who's on First and OpenStreetMap often have their own `bbox` elements. This `bbox` is at the same level as `properties`. If present, it describes the geographic extent of the feature, such as the screen size necessary to show all of California without needing to send the precise polygon geometry. This should be treated as separate from the `bbox` that describes the entire `FeatureCollection`.
## Result count
@ -103,8 +107,8 @@ By default, Mapzen Search results 10 places, unless otherwise specified. If you
@ -14,9 +14,9 @@ With reverse geocoding with Mapzen Search, you can look up all sorts of informat
To get started with reverse geocoding, you need a latitude, longitude pair in decimal degrees specified with the parameters `point.lat` and `point.lon`, respectively. For example, the Eiffel Tower in Paris, France, is located at `48.858268,2.294471`. The reverse geocode query for this would be:
`point.lat` | floating point number | yes | none | `48.858268`
`point.lon` | floating point number | yes | none | `2.294471`
`size` | integer | no | `10` | `3`
@ -36,7 +36,7 @@ Parameter | Type | Required | Default | Example
A basic parameter for filtering is `size`, which is used to limit the number of results returned. In the earlier request that returned the Eiffel Tower (or 'Tour Eiffel', to be exact), notice that other results were returned including "Bureau de Gustave Eiffel" (a museum) and "Le Jules Verne" (a restaurant). To limit a reverse geocode to only the first result, pass the `size` parameter:
The default value for `size` is `10` and the maximum value is `40`. Specifying a value greater than `40` will override to `40` and return a warning in the response metadata.
@ -51,13 +51,13 @@ By default, reverse geocoding returns results from any [data source](data-source
| [Who's on First](https://whosonfirst.mapzen.com) | `whosonfirst` | `wof` |
Without specifying further, reverse geocoding doesn't restrict results to a particular type (street, venue, neighbourhood, and so on). If your application is only concerned with, say, which city a latitude, longitude is closest to, then use the `layers` parameter. For example, the following request returns only results that are localities (cities and towns):
Here are all the supported layers and their meanings.
@ -81,11 +81,11 @@ Here are all the supported layers and their meanings.
If you are performing a reverse geocode near a country boundary, and are only interested in results from one country and not the other, you can specify a country code. You can set the `boundary.country` parameter value to the alpha-2 or alpha-3 [ISO-3166 country code](https://en.wikipedia.org/wiki/ISO_3166-1). For example, the latitude,longitude pair `47.270521,9.530846` is on the boundary of Austria, Liechtenstein, and Switzerland. Without specifying a `boundary.country`, the first 10 results returned may come from all three countries. By including `boundary.country=LIE`, all 10 results will be from Liechtenstein. Here's the request in action:
Note that `UK` is not a valid ISO 3166-1 alpha-2 country code.
## Distance Confidence scores for the results
## Distance and confidence scores for the results
Each result returned has a distance from the query point (in meters) and an associated confidence score. Confidence scores are calculated based on the distance from the result to the supplied `point.lat` and `point.lon`. Confidence scoring for reverse geocode results is likely to change with different data sources and layers.
@ -103,16 +103,16 @@ Distance from `point.lat`/`point.lon` | Confidence score
This section shows how the various parameters can be combined to form complex use cases.
scheme domain version path query authentication token
```
@ -27,7 +27,7 @@ In the simplest search, you can provide only one parameter, the text you want to
For example, if you want to find a [YMCA](https://en.wikipedia.org/wiki/YMCA) facility, here's what you'd need to append to the base URL of the service, `search.mapzen.com`.
@ -55,7 +55,7 @@ In the example above, you will find the name of each matched locations in a prop
Spelling matters, but not capitalization when performing a query with Mapzen Search. You can type `ymca`, `YMCA`, or even `yMcA`. See for yourself by comparing the results of the earlier search to the following:
Note that the results are spread out throughout the world because you have not given your current location or provided any other geographic context in which to search.
@ -88,11 +88,11 @@ By default, Mapzen Search results 10 places, unless otherwise specified. If you
@ -106,7 +106,7 @@ Sometimes your work might require that all the search results be from a particul
Now, you want to search for YMCA again, but this time only in Great Britain. To do this, you will need to know that the alpha-3 code for Great Britain is GBR and set the parameters like this:
@ -184,7 +184,7 @@ Sometimes you don't have a rectangle to work with, but rather you have a point o
In this example, you want to find all YMCA locations within a 35-kilometer radius of a location in Ontario, Canada. This time, you can use the `boundary.circle.*` parameter group, where `boundary.circle.lat` and `boundary.circle.lon` is your location in Ontario and `boundary.circle.radius` is the acceptable distance from that location. Note that the `boundary.circle.radius` parameter is always specified in kilometers.
@ -243,13 +243,14 @@ Looking at the results, you can see that the few locations closer to this locati
* YMCA, Loughborough, United Kingdom [distance: 16978.367]
## Combine boundary search and prioritization
Now that you have seen how to use boundary and focus to narrow and sort your results, you can examine a few scenarios where they work well together.
### Prioritize within a country
Going back to the YMCA search you conducted with a focus around a point in Sydney, the results came back from distant parts of the world, as expected. But say you wanted to only see results from the country in which your focus point lies. You can combine that same focus point in Sydney with the country boundary of Australia like this.
@ -276,7 +277,7 @@ The results below look different from the ones you saw before with only a focus
If you are looking for the nearest YMCA locations, and are willing to travel no farther than 50 kilometers from your current location, you likely would want the results to be sorted by distance from current location to make your selection process easier. You can get this behavior by using `focus.point` in combination with `boundary.circle.*`. You can use the `focus.point.*` values as the `boundary.circle.lat` and `boundary.circle.lon`, and add the required `boundary.circle.radius` value in kilometers.
@ -318,7 +319,7 @@ The search examples so far have returned a mix of results from all the data sour
If you use the `sources` parameter, you can choose which of these data sources to include in your search. So if you're only interested in finding a YMCA in data from OpenAddresses, for example, you can build a query specifying that data source.
@ -341,7 +342,7 @@ Because OpenAddresses is, as the name suggests, only address data, here's what y
If you wanted to combine several data sources together, set `sources` to a comma separated list of desired source names. Note that the order of the comma separated values does not impact sorting order of the results; they are still sorted based on the linguistic match quality to `text` and distance from `focus`, if you specified one.
* [Alb-Donau-Kreis](https://mapzen.github.io/search-sandbox/?query=search/structured&county=Alb-Donau-Kreis&country=DEU) in Germany
Counties are not as commonly used in geocoding as localities, but can be useful when attempting to disambiguate between localities. For instance, there are three cities named Red Lion in Pennsylvania but only one in each of three counties. Specifying a county disambiguates this list to a single result.
@ -99,9 +99,9 @@ Counties are not as commonly used in geocoding as localities, but can be useful
Regions in the United States have [common abbreviations](https://en.wikipedia.org/wiki/List_of_U.S._state_abbreviations), such as PA for [Pennsylvania](https://whosonfirst.mapzen.com/spelunker/id/85688481/) and NM for [New Mexico](https://whosonfirst.mapzen.com/spelunker/id/85688493/). The `region` parameter can be a full name or abbreviation, so specifying `/v1/search/structured?region=NM` is functionality equivalent to `/v1/search/structured?region=New Mexico`.
Keep in mind that you can search for `postalcode` exclusively. So requests like [/v1/search/structured?postalcode=87801](http://search.mapzen.com//v1/search/structured?postalcode=87801) will return matching postalcode records.
Keep in mind that you can search for `postalcode` exclusively. So requests like [/v1/search/structured?postalcode=87801]( https://mapzen.github.io/search-sandbox/?query=search/structured&postalcode=87801) will return matching postalcode records.
### country
@ -123,9 +123,9 @@ Keep in mind that you can search for `postalcode` exclusively. So requests like
For example, [Peach Bottom, Pennsylvania](https://whosonfirst.mapzen.com/spelunker/id/404487863/) is only a `localadmin` place type and not a `locality` in Who's on First. For simplicity, if a structured geocoding request specifies `locality=Peach+Bottom®ion=Pennsylvania`, then `Peach Bottom` in both the `locality` and `localadmin` layers are searched.
For example, [Peach Bottom, Pennsylvania](https://whosonfirst.mapzen.com/spelunker/id/404487863/) is only a `localadmin` place type and not a `locality` in Who's on First. For simplicity, if a structured geocoding request specifies `locality=PeachBottom®ion=Pennsylvania`, then `Peach Bottom` in both the `locality` and `localadmin` layers are searched.
Rhonda Glennon
8 years ago
committed by
GitHub