A modular, open-source search engine for our world.

Pelias is a geocoder powered completely by open data, available freely to everyone.

Local Installation · Cloud Webservice · Documentation · Community Chat

What is Pelias?

Pelias is a search engine for places worldwide, powered by open data. It turns addresses and place names into geographic coordinates, and turns geographic coordinates into places and addresses. With Pelias, you’re able to turn your users’ place searches into actionable geodata and transform your geodata into real places.

We think open data, open source, and open strategy win over proprietary solutions at any part of the stack and we want to ensure the services we offer are in line with that vision. We believe that an open geocoder improves over the long-term only if the community can incorporate truly representative local knowledge.

Pelias CSV Importer

This importer is designed to bring data into Pelias from a properly formatted CSV file.

It's originally based off of the OpenAddresses importer, which also uses a CSV format.

Overview

This importer will process any CSV, attempting to create a Pelias document for each row.

In order to be useful, each row needs to define a source, a latitude, a longitude, and a name. Address components can optionally be specified.

This importer will accept any column name as uppercase or lowercase. Lowercase has priority if both are present.

Latitude

Latitude can come from a column called lat. It should be a WGS84 value between -90.0 and 90.0.

Longitude

Longitude can come from a column called lon. It should be a WGS84 value between -180.0 and 180.0.

Address

A valid address consists of at least a street, and possibly a housenumber and postalcode.

Valid column names for street are: street

Valid column names for housenumber are: housenumber, number

Valid column names for postalcode are: postalcode, postcode, zipcode

Valid column names for intersections are: cross_street (note: street is also required!)

Name

A free-form string that represents the name of a record. It might be the name of a venue which also has an address, or the name of a city, mountain, or other interesting feature.

Valid column names for name are: name.

Layer

Pelias allows sorting records into different layers, representing different classes of data.

The most common layers are address, street, and venue. Address and street have special meaning to Pelias: when Pelias looks for an address, it can also attempt to use its interpolation engine to fill in missing addresses. If no addresses (exact or interpolated) are found, Pelias will try to find a street record matching the street from the original address in the query.

Another type of layer is "administrative" layers such as city and country.

Layers do not have to fall into these categories. Any layer that doesn't have special meaning to Pelias can still be use to filter with the layers parameter to the Pelias API.

Valid column names for the layer value are: layer and layer_id

Source

Pelias understands that different data records come from different sources, and allows filtering based on source. Common data projects that represent sources in Pelias include OpenStreetMap, OpenAddresses, and Who's on First.

Custom data with arbitrary sources are supported by this importer and can be used for user filtering. The source value won't have any other effect on how Pelias treats a record when querying.

Valid column names for the source value are : source

ID

An ID is a unique identifier for each record. Pelias IDs are strings, so they can contain text. Pelias records must have a unique source, layer, and ID. Attempting to create multiple records with the same source, layer and ID will cause all but the most recent record to be overwritten.

If an ID is not specified for a row in a CSV, the row number will be used.

Names in multiple Languages

Multiple names in different languages can be assigned by using the name_$lang fields, where $lang is an ISO 639-1 language code.

For example, to create a record for London in English and French, use the following CSV:

id	name	name_fr	source	layer	lat	lon
1	London	Londres	custom	locality	5	6

Multiple alias names

A record can have multiple aliases, or alternative names, specified as an array using the name_json field.

The following CSV will create a record for John F Kennedy International Airport, with common aliases including JFK and JFK airport.

id	name	name_json	source	layer	lat	lon
1	John F Kennedy International Airport	"[""JFK"", ""JFK Airport""]"	custom	venue	40.639722	-73.778889

The contents of the name_json field must be a JSON array. As a reminder, in CSV files, records that contain commas must be quoted using double quotes, and records with a double quote in the value itself must be double-double-quoted, as shown above.

Aliases and languages can both be specified. For example, the name_json_es field allows setting multiple aliases in Spanish.

Popularity

Popularity values can be specified to mark records as more important than others. This value should be an integer greater than zero, in the popularity column.

Categories

Category values can be added to a record. For a single category, use the category field. For multiple categories, use category_json, with the same formatting as for alias names.

Parent

Parent information for record can be added using the parent_json field.

Only the valid parent field names specified in the pelias/model are supported, records with parent containing unsupported fields names will be ignored with a warning in the logs and will not be imported.

List of valid fields in pelias/model, which eventually should match the list of valid fields in pelias/schema.

The contents of the parent_json field must be a valid JSON object. An example of the valid contents of parent_json field are:

    {
	"county": [{
		"id": "34",
		"name": "Innlandet",
		"abbr": "InL",
		"source": "OSM",
	}],
	"country": [{
		"id": "NOR",
		"name": "Norway"
		"abbr": "NO"
	}],
	"locality": [{
		"id": "3403",
		"name": "Hamar"
		"source": "SomeSource"
	}]
    }

In CSV files, records that contain commas must be quoted using double quotes, and records with a double quote in the value itself must be double-double-quoted, as shown below in the example for the parent_json field.

"{""county"":[{""id"":""34"",""name"":""Innlandet""}],""country"":[{""id"":""NOR"",""name"":""NO""}],""locality"":[{""id"":""3403"",""name"":""Hamar""}]}"

The valid properties for any parent field are id, name, abbr (abbreviation), source, where id and name are mandatory fields. Any other fields will be ignored without any warning. In case the mandatory fields are missing the record will be ignored with the warning in the logs and will not be imported.

In the case where multiple parent values are provided for the same field name, we store all copies in the elastic index, making them all searchable, but only the first entry is used for displaying the label.

Custom data

Arbitrary custom data that does not fit into the standard Pelias schema can be stored for later retrieval under the addendum property.

Currently, custom data is supported when encoded as any valid JSON object. In the future, support for adding individual values via CSV columns will be supported.

Custom data entires are namespaced, so this importer supports any column starting with addendum_json_. The rest of the column name will determine the namespace.

For example, to store a WikiData and Geonames concordance ID, the following CSV format might be used:

id	name	source	layer	lat	lon	addendum_json_geonames	addendum_json_wikidata
1	test	custom	venue	5	6	"{ ""id"": 600 }	{ ""id"": ""Q47"" }"

The Pelias API will then return a GeoJSON Feature like the following:

{
  "properties": {
    "id": "1",
    "gid": "custom:venue:1",
    "layer": "venue",
    "source": "custom",
    "source_id": "1",
    "name": "test",
    "confidence": 1,
    "match_type": "exact",
    "accuracy": "centroid",
    "label": "London, England, United Kingdom",
    "addendum": {
      "geonames": {
        "id": 600
      },
      "wikidata": {
        "id": "Q47"
      }
    }
  }
}

Requirements

Node.js is required.

See Pelias software requirements for supported versions.

Installation

git clone https://github.com/pelias/csv-importer
cd csv-importer
npm install

Usage

# download files, if desired
./bin/download

# run an import
./bin/start

Downloading CSV files

This importer includes a downloader that supports downloading any uncompressed CSV files over HTTP/HTTPS.

Configuration

This importer can be configured in pelias-config, in the imports.csv hash. A sample configuration file might look like:

{
  "esclient": {
    "hosts": [
      {
        "env": "development",
        "protocol": "http",
        "host": "localhost",
        "port": 9200
      }
    ]
  },
  "api": {
    "targets": {
      "yoursource": ["yourlayers"]
    }
  },
  "logger": {
    "level": "debug"
  },
  "imports": {
    "whosonfirst": {
      "datapath": "/mnt/data/whosonfirst/",
      "importPostalcodes": false,
      "importVenues": false
    },
    "csv": {
      "datapath": "/path/to/your/csv/files",
      "files": [],
      "download": [
        "https://example.com/csv-to-download.csv"
      ]
    }
  }
}

Important: You must put any custom source and layers imported by your data in pelias.json as explained in the relevant API configuration documentation. For a reasonably common use case for the source csv with only records in the address layer, the following configuration is a good starting point:

{
  "api": {
    "targets": {
      "csv": ["address"]
    }
  }
}

The following properties are recognized:

This importer is configured using the pelias-config module. The following configuration options are supported by this importer.

key	required	default	description
datapath	yes		The absolute path of the directory containing data files, or where downloaded files will be stored.
files	no	[]	An array of the names of the files to import. If specified, only these files will be imported. If not specified, or empty, all .csv files in the given directory will be imported.
download	no	[]	An array of URLs of CSV files that can be downloaded. Files must be plain-text (uncompressed) CSV files