Scraper development XML
|Development||Add-on development||Scraper development XML|
|This page or section may require cleanup, updating, spellchecking, reformatting and/or updated images. Please improve this page if you can. The discussion page may contain suggestions.|
Scrapers (also outside the Kodi community more commonly referred to as "web scrapers" or "importers") are used by Kodi to go online to automatically retrieve metadata information and artwork from websites for Music, Videos, Movies and TV Shows. This metadata information and artwork is then used for Kodi's music and video libraries.
1 Introduction to Scrapers
Before version 2.0 of Kodi Media Center the only available web scraper we had was a IMDb importer. In later versions it has become possible to create custom scrapers to collect data from almost any web-page which list information in a consistent way, and today Kodi comes with many scrapers for several languages and different websites.
To learn more about actually using scrapers in Kodi, please look at:
And to learn more about creating scrapers, please look at this article:
Kodi come with several scrapers for Movies, TV shows and Music Videos which are stored in xbmc\system\scrapers\video. They are just specially formatted XML files.
- The location of the scrapers has changed for EDEN Beta 3 - the \scrapers directory is old. They are now add-ons- see the 'the movieDB' as an example- just copy the directories metadata.themoviedb.org and metadata.common.themoviedb.org to your own names, change a few xml guts and you will have a new scraper that you can play with. THEN these instructions become relevant again.
The scraper XML file consists of text processing operations that work over a set of text buffers, labelled $$1 to $$20. The basic operation is RegExp. A single RegExp is a text processing command that takes text in buffer 'input', runs <expression> on it... formats the matches with 'output' and places (or appends) the result in buffer 'dest'. A scraper XML file must use RegExps to implement functions that Kodi will call such as GetSearchResults.
- Knowledge of regular expressions
- Basic XML syntax knowledge.
To see a full scraper, see the themoviedb reference implementation in GIT. The general layout of scraper.xml is as follows:
<scraper> <NfoUrl> <RegExp> <expression></expression> </RegExp> </NfoUrl> <CreateSearchUrl> <RegExp> <expression></expression> </RegExp> </CreateSearchUrl> <GetSearchResults> <RegExp> <expression></expression> </RegExp> </GetSearchResults> <GetDetails> <RegExp> <expression></expression> </RegExp> </GetDetails> <CustomFunction> <RegExp> <expression></expression> </RegExp> </CustomFunction> </scraper>
If RegExp tags are being nested they are being worked through in a lifo manner.
1.3 Kodi/Scraper Interaction
1.4 XML character entity references
Unlike traditional HTML with its large range of character entity references, in XML there are only five predefined character entity references. These are used to escape characters that are markup sensitive in certain contexts:
- & → &
- < → <
- > → >
- " → "
- ' → '
This means that in the regular expressions you actually have to use entities instead of the actual characters.
For example, the following would be wrong:
<CreateSearchUrl dest="3"> <RegExp input="$$1" output="http://www.example.com/search.php?title=\1&new=true" dest="3"> <expression noclean="1"> <a href="example">.[^<]* </expression> </RegExp> </CreateSearchUrl>
<CreateSearchUrl dest="3"> <RegExp input="$$1" output="http://www.example.com/search.php?title=\1&amp;new=true" dest="3"> <expression noclean="1"> &lt;a href=&quot;example&quot;&gt;.[^&lt;]* </expression> </RegExp> </CreateSearchUrl>
If you open the scraper XML file in a web browser the actual characters are shown for the entities, be careful if you cut&paste.
1.5 Regular Expression Engine
There are a few things to note about the regular expression engine:
- Laziness doesn't work. e.g. <xml>.*?</xml> (For a workaround see http://www.regular-expressions.info/repeat.html An Alternative to Laziness)
- \w or \d does not work, use [a-zA-Z] and [0-9] instead
- Regular expressions are case sensitive
- A dot matches a new line
- Beginning of line ^ and end of line $ does not seem to work
1.6 Tags in Detail
<scraper name="jaded video" content="movies" thumb="jaded.jpg" framework="1.0" date="2009-05-22"></scraper>
The sole purpose of this tag is to define the name, type and thumbnail of scraper.
- name: The name of the scraper
- content: movies/tvshows/mvid
- thumb: relative path to a scraper thumbnail
- framework: Scraper Framework being used. The latest is 1.0 (as used in Kodi 9.04 Babylon)
- date: Date this version of the scraper has been released in ISO 8601
This is an output string that the scraper.xml generates and passes to Kodi. Url tags can have the following attributes:
- spoof: The referrer url which should be send. This is sometimes needed for pages with direct linking protection.
- post: if the post attribute is present variables in the url will be send via POST.
- function: See GetDetails.
Every url returned from the scrapers has to be inside an url tag. An exception is CreateSearchUrl where the url tag is optional.
A RegExp is a text processing command that takes text in buffer 'input', runs <expression> on it... formats the matches with 'output' and places the result in buffer 'dest'
<RegExp conditional="override" input="$$5" output="<details>\1</details>" dest="3"> <expression repeat="yes" noclean="1" trim="1" clear="no">([a-zA-Z][^,]*)</expression> </RegExp>
The <expression> tag holds the regular expression. Here it is: <xml>([a-zA-Z][^,]*)whatever (.[^-]*)</xml> The regular expression matches are stored in \1, \2, ... , \9 and can be used in the output attribute of the RegExp tag. If the expression is blank, the input buffer contents is put into match \1.
- repeat="yes/no": Repeat the regular expression.
- noclean="1,..,9": By default html tags and special characters are stripped from the matches \1, ..., \9. By setting noclean="1, ..., 9" you can stop this behavior.
- trim="1,..,9": Trim white spaces from the end of matches 1 to 9.
- clear=yes/no": If set to yes, if the expression fails dest is cleared
The RegExp tag sets the input which will is being searched by the regular expression
- input="$$x" where x=1 to 9: The input variable holds the text that will get searched by the regular expression
- output: Defines how the output should look like. Here you can use \1, ..., \9 which represent regular expression matches.
- dest="x" where x=1 to 9: The variable to which the output should be stored to. If clear is true, its previous content will get cleared if the expression fails.
- conditonal="<condition>": A condition that must resolve to TRUE for the particular RegExp to be run. Currently the only available condition is "override", which is set based on the Language Override setting in the scraper.
It's the task of NfoUrl to return the url for GetDetails from an nfo.
- $$1: Variable 1 holds the content of the Nfo file.
<CreateSearchUrl dest="3"> <RegExp input="$$1" output="http://akas.imdb.com/find?s=tt;q=\1" dest="3"> <expression noclean="1"></expression> </RegExp> </CreateSearchUrl>
- $$1: Variable 1 holds the search string. This is usually the filename stripped by some words e.g. DVDRip, Xvid ..
The purpose of this function is to create a variable which will hold the url of the search result page.
- dest="x" where x=1 to 9: Variable x shall hold the url of the search result page.
In the example above <CreateSearchUrl dest="3"> means that variable 3 should hold the search results page url. The Regular expression tag searches through variable 1 and stores the searchstring in variable 3. Since no expression is specified the entire variable $$1 is matched and stored in \1. Since this does not contain html tags it is not necessary to clean \1.
<GetSearchResults dest="8"> <RegExp input="$$5" output="<?xml version="1.0" encoding="iso-8859-1" standalone="yes"?><results>\1</results>" dest="8"> <RegExp input="$$1" output="<entity><title>\1</title><url>http://jadedvideo.com/yz_resultJAVA.asp?PRODUCT_ID=\2</url></entity>" dest="5"> <expression repeat="yes"><font color="#000FF">(.[^"]*)</big>.[^"]*SKU: ([0-9]+)<br></expression> </RegExp> <expression noclean="1"></expression> </RegExp> </GetSearchResults>
- $$1: Variable 1 hold the content of the search URL returned to CreateSearchUrl.
It is the task of this function to return a list of search results in the following format:
<?xml version="1.0" encoding="iso-8859-1" standalone="yes"?> <results> <entity> <title></title> <url></url> <url></url> </entity> <entity> <title></title> <url></url> <url></url> </entity> </results>
There can be up to 9 <url> tags within each movie tag. The webpages specified in the url tags should hold detailed information about the movie/tv series/mvid. In most cases one url will be enough. The url tag can have the argument function="nameoffunction" e.g.
In this case other then GetDetails, "CustomFunction" will be executed after GetDetails has run.
When using the scraper the user will be asked in Kodi to select one of the movies from the list which was returned by this function. The one which gets selected will be processed from there on.
By default Kodi will try to intelligently re-order the list of returned entries by comparing titles with the search string. If the results you are returning are already in a properly sorted order set <results sorted="yes">.
Here the actual data which is being saved in the database is being retrieved. For movies the following format should be returned:
<details> <title></title> <year></year> <director></director> <top250></top250> <mpaa></mpaa> <tagline></tagline> <runtime></runtime> <thumb></thumb> <credits></credits> <rating></rating> <votes></votes> <genre></genre> <actor> <name></name> <role></role> </actor> <outline></outline> <plot></plot> </details>
- $$x where x= 1 to 9: Variable x holds the content of the URL specified in the <url> tag number x in GetSearchResults.
e.g. if there are two <url> tags (without argument function) returned by GetSearchResults. Variable 1 will contain the content of the first url tag and variable 2 the content of the second url tag.
The thumb tag can have a spoof argument (like <url>). This feature can be used to circumvent direct linking protections. The firefox extensions refspoof http://refspoof.mozdev.org/ can help in testing if referrer spoofing works.
- The following tags can appear more than once in <details>:
This function should return an xml similar to GetDetails. For tags that can have multiple values the values of CustomFunction and GetDetails are combined (genre etc). For the rest the last CustomFunction will take precedence.
- $$1: Variable 1 holds the content of the URL specified in the url tag with the function="CustomFunction" argument.
variable 1 would hold the content of "http://www.imdb.com/title/tt0452624/".