This is a simple program for parsing a list of addresses using google maps api.
The main function for parsing a single address is implemented on src/main/scala/AddressParser.scala:
- It builds the proper URL google maps query with the requested address
- It uses Play Json to parse the json response
- It extracts the data we need and it builds an Address case class
BatchParsderCmd queries the addresses from a database, queries each address using google maps api, parses it and saves the result to the DB.
It is implemented in Scala with akka streams and slick for optimal performance.
You need to install:
- Java JDK 8
- SBT (build tool for Scala)
- Google Maps Geocoding API key. Get your free account here.
It will be something like
AIzaSyBwG-Zo6me1yd6V2_ZO7L-3K8A0U1122LA - Optional: IntelliJ IDE with the Scala plugin
$ export googleApiKey="AIzaSyBwG-Zo6me1yd6V2_ZO7L-3K8A0U1122LA" # update with your valid api key
$ sbt "runMain ParserCmd $googleApiKey \"Av du Rond-Point 1, Lausanne, Switzerland\""
- parsedAddress:
| - numResults: 2
| - addressComponents: # only are take into account the ones defined on src/main/scala/AddressParser.scala -> addressComponentTypes
| | - country: United States
| | - administrative_area_level_1: West Virginia
| | - administrative_area_level_2: Putnam County
| | - postal_code: 25109
| | - route: 1st Avenue North
| | - locality: Hometown
| | - administrative_area_level_3: Buffalo-Union
| - location:
| | - lat: 38.535866
| | - lng: -81.86641
| - formattedAddress: 1st Ave N, Hometown, WV 25109, USA
| - mainType: route # this is computed according to src/main/scala/AddressParser.scala -> mainTypeOrder
| - types: [route]
| - viewportArea: 361119.3987508803 # this is computed from result.geometry.viewport
Given this example address: "Av du Rond-Point 1, Lausanne, Switzerland"
We construct this google query url (replace your valid api key):
https://maps.googleapis.com/maps/api/geocode/json?address=Av%20du%20Rond-Point%201,%20Lausanne,%20Switzerland&key=AIzaSyBwG-Zo6me1yd6V2_ZO7L-3K8A0U1122LA
We get this JSON response:
{
"results" : [
{
"address_components" : [
{
"long_name" : "81",
"short_name" : "81",
"types" : [ "street_number" ]
},
{
"long_name" : "Rue de Genève",
"short_name" : "Rue de Genève",
"types" : [ "route" ]
},
{
"long_name" : "Lausanne",
"short_name" : "Lausanne",
"types" : [ "locality", "political" ]
},
{
"long_name" : "Lausanne",
"short_name" : "Lausanne",
"types" : [ "administrative_area_level_2", "political" ]
},
{
"long_name" : "Vaud",
"short_name" : "VD",
"types" : [ "administrative_area_level_1", "political" ]
},
{
"long_name" : "Switzerland",
"short_name" : "CH",
"types" : [ "country", "political" ]
},
{
"long_name" : "1004",
"short_name" : "1004",
"types" : [ "postal_code" ]
}
],
"formatted_address" : "Rue de Genève 81, 1004 Lausanne, Switzerland",
"geometry" : {
"bounds" : {
"northeast" : {
"lat" : 46.5249294,
"lng" : 6.613929199999999
},
"southwest" : {
"lat" : 46.52491699999999,
"lng" : 6.613923799999999
}
},
"location" : {
"lat" : 46.5249294,
"lng" : 6.613929199999999
},
"location_type" : "RANGE_INTERPOLATED",
"viewport" : {
"northeast" : {
"lat" : 46.52627218029149,
"lng" : 6.615275480291501
},
"southwest" : {
"lat" : 46.5235742197085,
"lng" : 6.612577519708497
}
}
},
"place_id" : "EihSdWUgZGUgR2Vuw6h2ZSA4MSwgMTAwNCBMYXVzYW5uZSwgU3Vpc3Nl",
"types" : [ "street_address" ]
}
],
"status" : "OK"
}
It parses the json response using the play json library.
It uses the implicit formats, which use the name of the case class fields to extract the data from the json response, and it ignores the rest of the json response.
Then it extracts the data of interest (for instance, it finds the address_component with at least one type "administrative_area_level_1", and it maps it to areaLevel1).
It builds an Address case class (defined also at AddressParser.scala) as shown in the previous "Example: parse one address" section.
This was enough for our use case, but it is easy to modify AddressParser.scala for extracting more info from the json response.
We have a sql database, with a list of addresses to query and a place to save the extracted information.
We have a table called addresses, and it contains these fields:
- unformattedAddress
- googleResponse
- parseGoogleResponseStatus
- numResults
- locality
- administrative_area_level_1
- administrative_area_level_2
...
We can create the table as follows:
mysql>
create database test default character set utf8mb4 collate utf8mb4_unicode_ci;
use test
drop table if exists addresses;
create table addresses (
id int not null auto_increment primary key,
unformattedAddress varchar(500) not null,
ts timestamp default current_timestamp on update current_timestamp,
googleResponse longtext,
parseGoogleResponseStatus longtext,
numResults int,
formattedAddress varchar(500),
lat float(10,6), lng float(10,6), mainType varchar(200), types longtext, viewportArea float,
administrative_area_level_1 varchar(200), administrative_area_level_2 varchar(200), administrative_area_level_3 varchar(200), administrative_area_level_4 varchar(200), administrative_area_level_5 varchar(200), airport varchar(200), country varchar(200), establishment varchar(200), floor varchar(200), locality varchar(200), natural_feature varchar(200), neighborhood varchar(200), park varchar(200), point_of_interest varchar(200), post_box varchar(200), postal_code varchar(200), postal_code_prefix varchar(200), postal_code_suffix varchar(200), postal_town varchar(200), premise varchar(200), route varchar(200), street_address varchar(200), street_number varchar(200), sublocality varchar(200), sublocality_level_1 varchar(200), sublocality_level_2 varchar(200), sublocality_level_3 varchar(200), sublocality_level_4 varchar(200), sublocality_level_5 varchar(200), subpremise varchar(200), ward varchar(200),
unique index(unformattedAddress), index(ts), index(googleResponse(100)), index(parseGoogleResponseStatus(100)), index(numResults), index(formattedAddress),
index(lat), index(lng), index(mainType), index(types(100)), index(viewportArea),
index(administrative_area_level_1), index(administrative_area_level_2), index(administrative_area_level_3), index(administrative_area_level_4), index(administrative_area_level_5), index(airport), index(country), index(establishment), index(floor), index(locality), index(natural_feature), index(neighborhood), index(park), index(point_of_interest), index(post_box), index(postal_code), index(postal_code_prefix), index(postal_code_suffix), index(postal_town), index(premise), index(route), index(street_address), index(street_number), index(sublocality), index(sublocality_level_1), index(sublocality_level_2), index(sublocality_level_3), index(sublocality_level_4), index(sublocality_level_5), index(subpremise), index(ward)
) engine = InnoDB default character set = utf8mb4 collate = utf8mb4_bin row_format=dynamic;
Note: we use float(10,6) for storing lat and lng, as proposed by the google api example.
Note: if you get an ERROR 1709 (HY000): Index column size too large. The maximum column size is 767 bytes., make sure you added row_format=dynamic at the end of the query, or configure correctly mysql server for row_format dynamic, or update your mysql server to the latest version.
Note: if you get an ERROR 1713 (HY000): Undo log record is too big., try updating the table with alter table addresses row_format=redundant;
Note: read about collate = utf8mb4_bin vs collate = utf8mb4_unicode_ci;
We can insert the addresses to query as follows:
mysql> insert into addresses_todelete1 (unformattedAddress) values
('Av du Rond-Point 1, Lausanne, Switzerland'),
('Statue, Palayam, Thiruvananthapuram, Kerala 695001, India'),
('Frankfurt am Main 80, DE'),
('(GSF);;INGOLSTAEDTER LANDSTR. 1;;D-8042 NEUHERBERG, DE'),
('CAREL VAN BYLANDTLAAN 30, LA HAYA NL, NL');
or from another table/query, such as from patstat:
mysql> insert ignore into addresses (unformattedAddress) select distinct(trim(concat(address_freeform, ', ', person_ctry_code))) from tls226_person_orig where nullif(address_freeform, '') is not null;
To connect to the database, we need to specify a JDBC URL. It looks as follows, depending on whether it is mysql, postgresql or something else:
$ export dbUrl="jdbc:mysql://SERVER_HOST/DATABASE?user=USER&password=PASSWORD&useSSL=false&useUnicode=yes&characterEncoding=utf8"
$ export dbUrl="jdbc:postgresql://SERVER_HOST/DATABASE?user=USER&password=PASSWORD&ssl=true"
$ export googleApiKey="AIzaSyBwG-Zo6me1yd6V2_ZO7L-3K8A0U1122LA" # update with your valid api key
$ sbt "runMain BatchParserCmd --help"
Usage: BatchParserCmd [options]
--op <value> where value = googleQueryAndParse, googleQueryOnly or parseOnly
--maxEntries <value>
--maxGoogleAPIRequestsPerSecond <value>
--maxGoogleAPIFatalErrors <value>
--googleApiKey <value>
--dbUrl <value>
--tableName <value>
--version
$ sbt "runMain BatchParserCmd --op=googleQueryAndParse --maxEntries=20 --maxGoogleAPIRequestsPerSecond=10 --maxGoogleAPIFatalErrors=5 --googleApiKey="$googleApiKey" --dbUrl="$dbUrl" --tableName=addresses"
maxEntries is the max number of google queries to do. It's best to try with a small number first.
The program will also stop if the max number of queries to the google api is exceeded (2500 request per day for the free account).
You can then execute this same command the following day, and it will resume the process (it will not re-download what it has queried already).
The program queries google in parallel. The bigger maxGoogleAPIRequestsPerSecond, the faster to query all addresses. Google has a rate limit and a daily limit, so try before with smalls numbers.
The program will stop after maxGoogleAPIFatalErrors. Set a small number.
mysql>
set @database = 'david'; set @table = 'addresses';
set @sql = concat('create view ', @table, '_ as select ', (select replace(group_concat(column_name), 'googleResponse,', "concat(left(replace(googleResponse, '\n', ' '), 20), '...') as googleResponseCrop,") from information_schema.columns where table_name = @table and table_schema = @database), ' from ', @table);
prepare stmt from @sql; execute stmt;
mysql> select * from addresses_;
mysql> select * from addresses_todelete1_;
+----+-----------------------------------------------------------+---------------------+-------------------------+---------------------------+------------+-----------------------------------------------------------+-----------+------------+---------------------+---------------------------------------------+--------------+-----------------------------+-----------------------------+-----------------------------+-----------------------------+-----------------------------+---------+---------------+---------------+-------+--------------------+-----------------+--------------+------+-------------------+----------+-------------+--------------------+--------------------+-------------+---------+-----------------------+----------------+---------------+-------------+---------------------+---------------------+---------------------+---------------------+---------------------+------------+------+
| id | unformattedAddress | ts | googleResponseCrop | parseGoogleResponseStatus | numResults | formattedAddress | lat | lng | mainType | types | viewportArea | administrative_area_level_1 | administrative_area_level_2 | administrative_area_level_3 | administrative_area_level_4 | administrative_area_level_5 | airport | country | establishment | floor | locality | natural_feature | neighborhood | park | point_of_interest | post_box | postal_code | postal_code_prefix | postal_code_suffix | postal_town | premise | route | street_address | street_number | sublocality | sublocality_level_1 | sublocality_level_2 | sublocality_level_3 | sublocality_level_4 | sublocality_level_5 | subpremise | ward |
+----+-----------------------------------------------------------+---------------------+-------------------------+---------------------------+------------+-----------------------------------------------------------+-----------+------------+---------------------+---------------------------------------------+--------------+-----------------------------+-----------------------------+-----------------------------+-----------------------------+-----------------------------+---------+---------------+---------------+-------+--------------------+-----------------+--------------+------+-------------------+----------+-------------+--------------------+--------------------+-------------+---------+-----------------------+----------------+---------------+-------------+---------------------+---------------------+---------------------+---------------------+---------------------+------------+------+
| 41 | Av du Rond-Point 1, Lausanne, Switzerland | 2017-11-20 18:02:08 | { "results" : [ ... | OK | 1 | Avenue du Rond-Point 1, 1006 Lausanne, Switzerland | 46.515224 | 6.628633 | point_of_interest | establishment, point_of_interest, premise | 61911.9 | Vaud | Lausanne | NULL | NULL | NULL | NULL | Switzerland | NULL | NULL | Lausanne | NULL | NULL | NULL | NULL | NULL | 1006 | NULL | NULL | NULL | NULL | Avenue du Rond-Point | NULL | 1 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL |
| 42 | Statue, Palayam, Thiruvananthapuram, Kerala 695001, India | 2017-11-20 18:02:08 | { "results" : [ ... | OK | 1 | Statue, Palayam, Thiruvananthapuram, Kerala 695001, India | 8.496704 | 76.950623 | sublocality_level_2 | political, sublocality, sublocality_level_2 | 386070 | Kerala | Thiruvananthapuram | NULL | NULL | NULL | NULL | India | NULL | NULL | Thiruvananthapuram | NULL | NULL | NULL | NULL | NULL | 695001 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | Statue | Palayam | Statue | NULL | NULL | NULL | NULL | NULL |
| 43 | Frankfurt am Main 80, DE | 2017-11-20 18:02:08 | { "results" : [ ... | OK | 3 | 80 Main Dr, Frankfort, IN 46041, USA | 40.296558 | -86.513023 | street_address | street_address | 68497.2 | Indiana | Clinton County | Center Township | NULL | NULL | NULL | United States | NULL | NULL | Frankfort | NULL | NULL | NULL | NULL | NULL | 46041 | NULL | NULL | NULL | NULL | Main Drive | NULL | 80 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL |
| 44 | (GSF);;INGOLSTAEDTER LANDSTR. 1;;D-8042 NEUHERBERG, DE | 2017-11-20 18:02:08 | { "results" : [],... | OK | 0 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL |
| 45 | CAREL VAN BYLANDTLAAN 30, LA HAYA NL, NL | 2017-11-20 18:02:08 | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL | NULL |
...
This is an example results after querying 250 addresses:
$ sbt "runMain BatchParserCmd --op=googleQueryAndParse --maxEntries=250 --googleApiKey="$googleApiKey" --dbUrl="$dbUrl" --tableName=addresses"
mysql> select count(*), googleResponse is not null as googleResponseStored, parseGoogleResponseStatus, numResults from addresses group by googleResponseStored, parseGoogleResponseStatus, numResults;
+----------+----------------------+-----------------------------------+------------+
| count(*) | googleResponseStored | parseGoogleResponseStatus | numResults |
+----------+----------------------+-----------------------------------+------------+
| 860838 | 0 | NULL | NULL |
| 58 | 1 | OK | 0 |
| 13 | 1 | OK | 2 |
| 179 | 1 | OK | 1 |
+----------+----------------------+-----------------------------------+------------+
4 rows in set (3.37 sec)
It says that:
- 179 cases are perfect (queried and parsed successfully).
- 13 cases, where google provided more than one result because the address is ambiguous (we take only the first result)
- 58 cases, where google did not find any result
- 860'838 cases that we still didn't query.
We can resume the program and query 100 more addresses:
$ sbt "runMain BatchParserCmd --op=googleQueryAndParse --maxEntries=100 --googleApiKey="$googleApiKey" --dbUrl="$dbUrl" --tableName=addresses"
mysql> select count(*), googleResponse is not null as googleResponseStored, parseGoogleResponseStatus, numResults from addresses group by googleResponseStored, parseGoogleResponseStatus, numResults;
+----------+----------------------+-----------------------------------+------------+
| count(*) | googleResponseStored | parseGoogleResponseStatus | numResults |
+----------+----------------------+-----------------------------------+------------+
| 860738 | 0 | NULL | NULL |
| 68 | 1 | OK | 0 |
| 19 | 1 | OK | 2 |
| 263 | 1 | OK | 1 |
+----------+----------------------+-----------------------------------+------------+
4 rows in set (3.67 sec)
and so on...
We keep a copy of the google responses in the database on the column googleResponse.
We do not need to re-query google in the case we change the code of AddressParser (to fix an issue, or to extract more data from the json response).
mysql> update addresses set parseGoogleResponseStatus = null, numResults = null;
Query OK, 274 rows affected (2.80 sec)
Rows matched: 861088 Changed: 274 Warnings: 0
mysql> select count(*), googleResponse is not null as googleResponseStored, parseGoogleResponseStatus, numResults from addresses group by googleResponseStored, parseGoogleResponseStatus, numResults;
+----------+----------------------+---------------------------+------------+
| count(*) | googleResponseStored | parseGoogleResponseStatus | numResults |
+----------+----------------------+---------------------------+------------+
| 860738 | 0 | NULL | NULL |
| 350 | 1 | NULL | NULL |
+----------+----------------------+---------------------------+------------+
2 rows in set (3.66 sec)
This tells us that we have the google response for 350 queries. Now we can re-execute the parser again for those 350 addresses (without querying google).
$ sbt "runMain BatchParserCmd --op=parseOnly --maxEntries=350 --dbUrl="$dbUrl" --tableName=addresses"
mysql> select count(*), googleResponse is not null as googleResponseStored, parseGoogleResponseStatus, numResults from addresses group by googleResponseStored, parseGoogleResponseStatus, numResults;
+----------+----------------------+-----------------------------------+------------+
| count(*) | googleResponseStored | parseGoogleResponseStatus | numResults |
+----------+----------------------+-----------------------------------+------------+
| 860738 | 0 | NULL | NULL |
| 68 | 1 | OK | 0 |
| 19 | 1 | OK | 2 |
| 263 | 1 | OK | 1 |
+----------+----------------------+-----------------------------------+------------+
4 rows in set (3.61 sec)
In case we want to force querying google again for any reason, we do as follows:
mysql> update addresses set googleResponse = null, parseGoogleResponseStatus = null, numResults = null;
$ sbt "runMain BatchParserCmd --op=googleQueryAndParse --maxEntries=250 --googleApiKey="$googleApiKey" --dbUrl="$dbUrl" --tableName=addresses"
In order to extract more data from the google json response, you need to
- modify the case classes
ResultandParsedAddress(onAddressParser.scala) - modify the function
parseAddressFromJsonResponse(googleResponseString: String): ParsedAddress(onAddressParser.scala) - modify the function
saveGoogleResponseToDatabase(unformattedAddress: String, googleResponse: String)(onBatchParserCmd.scala)