Elasticsearch River Web Plugin is a web crawler for Elasticsearch. This plugin provides a feature to crawl web sites and extract the content by CSS Query.
River Web | elasticsearch |
---|---|
master | 1.2.X |
1.1.2 | 1.1.1 |
1.1.1 | 1.0.2 |
1.0.1 | 0.90.7 |
Please file an issue.
River Web plugin depends on Quartz plugin. Quartz plugin needs to be installed before installing River Web plugin.
$ $ES_HOME/bin/plugin --install org.codelibs/elasticsearch-quartz/1.0.1
$ $ES_HOME/bin/plugin --install org.codelibs/elasticsearch-river-web/1.1.2
An index is needed to store crawl data before starting a river. For example, to store data to "webindex", create it as below:
$ curl -XPUT 'localhost:9200/webindex'
and then create a mapping setting if using "overwrite" option:
$ curl -XPUT "localhost:9200/webindex/my_web/_mapping" -d '
{
"my_web" : {
"dynamic_templates" : [
{
"url" : {
"match" : "url",
"mapping" : {
"type" : "string",
"store" : "yes",
"index" : "not_analyzed"
}
}
},
{
"method" : {
"match" : "method",
"mapping" : {
"type" : "string",
"store" : "yes",
"index" : "not_analyzed"
}
}
},
{
"charSet" : {
"match" : "charSet",
"mapping" : {
"type" : "string",
"store" : "yes",
"index" : "not_analyzed"
}
}
},
{
"mimeType" : {
"match" : "mimeType",
"mapping" : {
"type" : "string",
"store" : "yes",
"index" : "not_analyzed"
}
}
}
]
}
}'
"my_web" is a type given by your river name or "crawl.type".
A crawling configuration is created by registering a river as below. This example crawls sites of http://www.codelibs.org/ and http://fess.codelibs.org/ at 6:00am.
$ curl -XPUT 'localhost:9200/_river/my_web/_meta' -d '{
"type" : "web",
"crawl" : {
"index" : "webindex",
"url" : ["http://www.codelibs.org/", "http://fess.codelibs.org/"],
"includeFilter" : ["http://www.codelibs.org/.*", "http://fess.codelibs.org/.*"],
"maxDepth" : 3,
"maxAccessCount" : 100,
"numOfThread" : 5,
"interval" : 1000,
"target" : [
{
"pattern" : {
"url" : "http://www.codelibs.org/.*",
"mimeType" : "text/html"
},
"properties" : {
"title" : {
"text" : "title"
},
"body" : {
"text" : "body"
},
"bodyAsHtml" : {
"html" : "body"
},
"projects" : {
"text" : "ul.nav-list li a",
"isArray" : true
}
}
},
{
"pattern" : {
"url" : "http://fess.codelibs.org/.*",
"mimeType" : "text/html"
},
"properties" : {
"title" : {
"text" : "title"
},
"body" : {
"text" : "body",
"trimSpaces" : true
},
"menus" : {
"text" : "ul.nav-list li a",
"isArray" : true
}
}
}
]
},
"schedule" : {
"cron" : "0 0 6 * * ?"
}
}'
"my_web" is a configuration name for River, and you can replace it with one you want.
The configuration is:
Property | Type | Description |
---|---|---|
crawl.index | string | Stored index name. |
crawl.type | string | Stored type name. |
crawl.url | array | Start point of URL for crawling. |
crawl.includeFilter | array | White list of URL for crawling. |
crawl.excludeFilter | array | Black list of URL for crawling. |
crawl.maxDepth | int | Depth of crawling documents. |
crawl.maxAccessCount | int | The number of crawling documents. |
crawl.numOfThread | int | The number of crawler threads. |
crawl.interval | int | Interval time (ms) to crawl documents. |
crawl.incremental | boolean | Incremental crawling. |
crawl.overwrite | boolean | Delete documents of old duplicated url. |
crawl.userAgent | string | User-agent name when crawling. |
crawl.robotsTxt | boolean | If you want to ignore robots.txt, false. |
crawl.authentications | object | Specify BASIC/DIGEST/NTLM authentication info. |
crawl.target.urlPattern | string | URL pattern to extract contents by CSS Query. |
crawl.target.properties.name | string | "name" is used as a property name in the index. |
crawl.target.properties.name.text | string | CSS Query for the property value. |
crawl.target.properties.name.html | string | CSS Query for the property value. |
crawl.target.properties.name.script | string | Rewrite the property value by MVEL. |
schedule.cron | string | Cron format to start a crawler. |
If you want to stop the crawler, type as below: (replace my_web with your river name)
$ curl -XDELETE 'localhost:9200/_river/my_web/'
Full Text Search for Your site (ex. http://fess.codelibs.org/)
$ curl -XPUT 'localhost:9200/_river/fess/_meta' -d '{
"type" : "web",
"crawl" : {
"index" : "webindex",
"url" : ["http://fess.codelibs.org/"],
"includeFilter" : ["http://fess.codelibs.org/.*"],
"maxDepth" : 3,
"maxAccessCount" : 1000,
"numOfThread" : 5,
"interval" : 1000,
"target" : [{
"pattern" : {
"url" : "http://fess.codelibs.org/.*",
"mimeType" : "text/html"
},
"properties" : {
"title" : {
"text" : "title"
},
"body" : {
"text" : "body",
"trimSpaces" : true
}
}
}]
},
"schedule" : {
"cron" : "0 0 0 * * ?"
}
}'
$ curl -XPUT 'localhost:9200/_river/yahoo_com/_meta' -d '{
"type" : "web",
"crawl" : {
"index" : "webindex",
"url" : ["http://news.yahoo.com/"],
"includeFilter" : ["http://news.yahoo.com/.*"],
"maxDepth" : 1,
"maxAccessCount" : 10,
"numOfThread" : 3,
"interval" : 3000,
"userAgent" : "Mozilla/5.0 (Windows NT 6.3; Trident/7.0; rv:11.0) like Gecko",
"target" : [
{
"pattern" : {
"url" : "http://news.yahoo.com/video/.*html",
"mimeType" : "text/html"
},
"properties" : {
"title" : {
"text" : "title"
}
}
},
{
"pattern" : {
"url" : "http://news.yahoo.com/.*html",
"mimeType" : "text/html"
},
"properties" : {
"title" : {
"text" : "h1.headline"
},
"content" : {
"text" : "section#mediacontentstory p"
}
}
}
]
},
"schedule" : {
"cron" : "0 0 * * * ?"
}
}'
(if news.yahoo.com is updated, the above example needs to be updated.)
River Web supports BASIC/DIGEST/NTLM authentication. Set crawl.authentications object.
...
"numOfThread" : 5,
"interval" : 1000,
"authentications":[
{
"scope": {
"scheme":"BASIC"
},
"credentials": {
"username":"testuser",
"password":"secret"
}
}],
"target" : [
...
The configuration is:
Property | Type | Description |
---|---|---|
crawl.authentications.scope.scheme | string | BASIC, DIGEST or NTLM |
crawl.authentications.scope.host | string | (Optional)Target hostname. |
crawl.authentications.scope.port | int | (Optional)Port number. |
crawl.authentications.scope.realm | string | (Optional)Realm name. |
crawl.authentications.credentials.username | string | Username. |
crawl.authentications.credentials.password | string | Password. |
crawl.authentications.credentials.workstation | string | (Optional)Workstation for NTLM. |
crawl.authentications.credentials.domain | string | (Optional)Domain for NTLM. |
For example, if you want to use an user in ActiveDirectory, the configuration is below:
"authentications":[
{
"scope": {
"scheme":"NTLM"
},
"credentials": {
"domain":"your.ad.domain",
"username":"taro",
"password":"himitsu"
}
}],
River Web supports attachment type. For example, create a mapping with attachment type:
curl -XPUT "localhost:9200/web/test/_mapping?pretty" -d '{
"test" : {
"dynamic_templates" : [
{
...
"my_attachment" : {
"match" : "my_attachment",
"mapping" : {
"type" : "attachment",
"fields" : {
"file" : { "index" : "no" },
"title" : { "store" : "yes" },
"date" : { "store" : "yes" },
"author" : { "store" : "yes" },
"keywords" : { "store" : "yes" },
"content_type" : { "store" : "yes" },
"content_length" : { "store" : "yes" }
}
}
}
...
and then start your river. In "properties" object, when a value of "type" is "attachment", the crawled url is stored as base64-encoded data.
curl -XPUT 'localhost:9200/_river/test/_meta?pretty' -d '{
"type" : "web",
"crawl" : {
"index" : "web",
"url" : "http://...",
...
"target" : [
...
{
"settings" : {
"html" : false
},
"pattern" : {
"url" : "http://.../.*"
},
"properties" : {
"my_attachment" : {
"type" : "attachment"
}
}
}
]
...
An example in Japanese environment is below. First, put some configuration file into conf directory of Elasticsearch.
$ cd $ES_HOME/conf # ex. /etc/elasticsearch if using rpm package
$ sudo wget https://raw.github.com/codelibs/fess-server/master/src/tomcat/solr/core1/conf/mapping_ja.txt
$ sudo wget http://svn.apache.org/repos/asf/lucene/dev/trunk/solr/example/solr/collection1/conf/lang/stopwords_ja.txt
and then create "webindex" index with analyzers for Japanese. (If you want to use uni-gram, remove cjk_bigram in filter)
$ curl -XPUT "localhost:9200/webindex" -d '
{
"settings" : {
"analysis" : {
"analyzer" : {
"default" : {
"type" : "custom",
"char_filter" : ["mappingJa"],
"tokenizer" : "standard",
"filter" : ["word_delimiter", "lowercase", "cjk_width", "cjk_bigram"]
}
},
"char_filter" : {
"mappingJa": {
"type" : "mapping",
"mappings_path" : "mapping_ja.txt"
}
},
"filter" : {
"stopJa" : {
"type" : "stop",
"stopwords_path" : "stopwords_ja.txt"
}
}
}
}
}'
River Web allows you to rewrite crawled data by MVEL. In "properties" object, put "script" value to a property you want to rewrite.
...
"properties" : {
...
"flag" : {
"text" : "body",
"script" : "value.contains(\"Elasticsearch\") ? \"yes\" : \"no\""
},
The above is, if a string value of body element in HTML contains "Elasticsearch", set "yes" to "flag" property.
To start a crawler immediately, remove "cron" property in a configuration to register a river. No "cron" property means that the crawler starts right now and the river configuration is removed automatically at the end of the crawling.
Put "proxy" property in "crawl" property.
curl -XPUT 'localhost:9200/_river/my_web/_meta' -d '{
"type" : "web",
"crawl" : {
...
"proxy" : {
"host" : "proxy.server.com",
"port" : 8080
},
To set "isChildUrl" property to true, the property values is used as next crawled urls.
"crawl" : {
...
"target" : [
{
...
"properties" : {
"childUrl" : {
"value" : ["http://fess.codelibs.org/","http://fess.codelibs.org/ja/"],
"isArray" : true,
"isChildUrl" : true
},
You can insert your script to Starting River(start)/Executing Crawler(execute)/Finished Crawler(finish)/Closed River(close). To insert scripts, put "script" property to "crawl" property.
{
"crawl" : {
...
"script":{
"start":"your MVEL script...",
"execute":"your MVEL script...",
"finish":"your MVEL script...",
"close":"your MVEL script..."
},
River Web Plugin needs 'robot' index for web crawling. Therefore, in version 1.0.0 - 1.1.0, you need to create it before starting the crawl. Type the following commands to create 'robot' index:
$ curl -XPUT 'localhost:9200/robot/'
As of 1.1.1, "robot" index is created automatically.
In a river setting, "crawl.url" is starting urls to crawl a site, "crawl.includeFilter" filters urls whether are crawled or not, and "crawl.target.pattern.url" is a rule to store extracted web data. If a crawling url does not match "crawl.target.pattern.url", you would see the message. Therefore, it means the crawled url does not have an extraction rule.
For example, if you want to grab a content of description's meta tag, the configuration is below:
...
"target" : [
...
"properties" : {
...
"meta" : {
"attr" : "meta[name=description]",
"args" : [ "content" ]
},
"url" field needs to be "not_analyzed" in a mapping of your stored index. See Create Index To Store Crawl Data.
crawled data are stored to "robot" index during cralwing, data extracted from them are stored to your index specified by a river setting, and then data in "robot" index are removed when the crawler is finished.