How to customize Larbin

Where do the pages arrive ?

• void loaded (html *page) : This function is called when the fetch ended with success. From the page object, you can
  • get the url of the page by calling the method getUrl()
  • get the content of the page by calling the method getPage()
  • get the list of the sons by calling the method getLinks() (if options.h includes "#define LINKS_INFO")
  • get the http headers by calling the method getHeaders()
  • get the tag with getUrl()->tag (if options.h includes "#define URL_TAGS")
  For more details, see src/fetcher/file.h (for html), src/utils/url.h, src/utils/Vector.h.
• void failure (url *u, FetchError reason) : This function is called when the fetch ended by an error. u describes the url of the page. A description of its class can be found in src/utils/url.h. reason explains why the fetch failed. enum FetchError is defined in src/types.h.
• void initUserOutput () : Function for initialising all your data, called after all other initialisations
• void outputStats(int fds) : This function is called from the webserver if you want to track some data. fds is the file descriptor on which you must write to exchange with the net. This function is called in another thread than the main one with no lock at all, so be carefull !

For examples of useroutput files, see src/interf/xxxuseroutput.cc.

There are several default modules you can use. For more details, see options.h.

Simple customizations

larbin.conf

• From : YOUR mail : sent with http headers : very usefull when someone wants to complain about the robot :-(
• UserAgent : name of the robot (sent with each request)
• httpPort : port on which is launched the http statistic webserver (see http://localhost:8081/ when larbin is launched). If you set port to 0, no webserver will be launched. This can allow larbin not to launch a single thread.
• inputPort : port on which you can submit urls to fetch. If this line does not exist or if the port is 0, no input will be available.
• pagesConnexions : Number of page you fetch in parallel (to adapt depending of your network speed). Decrease this if you have too many timeouts (see stats) : 10% seems to be a maximum.
• dnsConnexions : Number of dns calls in parallel. 10 should be ok.
• depthInSite : How deep do you want to go in a site.
• noExternalLinks : Only follow links which are related to the same host.
• waitDuration : time between 2 calls at the same server in seconds. It should never be less than 30 s. However, even with 60 s, it won't slow the crawler much, and it is a much better behaviour.
• proxy : if you want to connect through a proxy (host port). Unless you have no other way to connect to the internet, you should not use this because it might slow the crawler a lot, and is probably also not so good for the proxy (especially if it has a cache).
• StartUrl : Where the search starts. This appears not to be very important, as soon as the page contains external urls.
• limitToDomain : with this option enabled, you will only crawl pages of some specific domain (.fr and .dk for example).
• forbiddenExtensions : What are the extensions you don't want ? (write all of them and terminate your list with end)

options.h

• The first thing you can define is the module you want to use for ouput. This defines what you want to do with the pages larbin gets. Here are the different options :
  • DEFAULT_OUTPUT : This module mainly does nothing, except statistics.
  • SIMPLE_SAVE : This module saves pages on disk. It stores 2000 files per directory (with an index).
  • MIRROR_SAVE : This module saves pages on disk with the hierarchy of the site they come from. It uses one directory per site.
  • STATS_OUTPUT : This modules makes some stats on the pages. In order to see the results, see http://localhost:8081/output.html.
  These modules can be customized in src/types.h.
  If you want to define a new module, please have a look at "src/interf/useroutput.cc", and do not hesitate to send me your work for inclusion.

• SPECIFICSEARCH : If this option is set, larbin's goal is to search for specific document. You must then define 2 arrays (NULL terminated) of char *, contentTypes and privilegedExts, which define respectively the content types which are looked for, and the extension of files (this extension is only used for speeding the search, pages are said to be specific only by looking at the content/type in http headers). You should also define another option telling how you want to manage specific pages :
  • DEFAULT_SPECIFIC : Default way of managing specific files : they are treated as html (ie same size limitations...), except that they are not parsed.
  • SAVE_SPECIFIC : Specific pages are saved on disk. this allows in particular specific pages to be much bigger (see src/types.h for customizating this module).
  • DYNAMIC_SPECIFIC : for big pages, larbin uses dynamically allocated buffers.
  If you want to define a new policy, please have a look at "src/fetch/specbuf.cc" and "src/fetch/specbuf.h", and do not hesitate to send me your work for inclusion.

• LINKS_INFO : Associate to each page the list of the links it contains. This information can be used in "useroutput.cc" with page->getLinks().
• FOLLOW_LINKS : If this option is not set, html pages won't be parsed and links won't be followed. This can be usefull when you feed larbin through the input system.
• NO_DUP : if this option is set, larbin does not return success when a page with the same content than an old one is encontered.
• URL_TAGS : if this option is set, an int is associated to every url (by default 0). If you use the input system, you'll have to give an int and the url instead of just the url. When the pages is fetched, you'll get it with the int (redirections are followed).
• EXIT_AT_END : If this option is set, larbin exits when there are not any more urls to get.
• IMAGES : If set, larbin gets the images contained in pages (ie follow img src links). Make sure to update forbiddenExtensions in larbin.conf according to your needs.
• ANYTYPE : If set, larbin gets every pages without caring about content type. Make sure to update forbiddenExtensions in larbin.conf according to your needs.
• COOKIES : If set, larbin manages cookies. Up to now, it is a very simple implementation, but it should be suitable in more than 90% of the situations.

• CGILEVEL : This option is foolowed by an integer which specified how reluctant to cgi you are. 0 means you want all cgis, 1 means you refuse urls with '?' or '=' inside, 2 means you also want to ban urls with 'cgi' inside.
• MAXBANDWIDTH : This option is followed by an integer which indicates the maximum bandwidth larbin should use. Because of the way bandwidth is limited, larbin might use 10 to 20 per cent more bandwidth than expected. If this option is not set, there is no bandwidth limitation.
• DEPTHBYSITE : If this option is set, when a links points to another site, the depth of the new url is reinitialized, else it is never.

• THREAD_OUTPUT : This option must be set if the code in "useroutput.cc" (the code you add) can use blocking instructions (read/write on network file descriptor...). If it is not set, there is only one thread in the program (except the webserver if any), so no locking is needed.
• RELOAD : If this option is enabled, larbin restarts from where it last stopped when you launch it. This allows to stop and restart larbin as needed (or restart after a crash). If you want to restart from scratch, use the -scratch option.

• NOWEBSERVER : Do not launch the webserver. This can be usefull if you don't want to launch any thread.
• GRAPH : Include nice histograms in the real time stat page.
• NDEBUG : Disable debugging information in the webserver.
• NOSTATS : Disable stats information in the webserver.
• STATS : Display stats on stdout every 8 seconds.
• BIGSTATS : Display the name of every page that is fetched on stdout. This might slow larbin quite much.
• CRASH : Should only be used for reporting terrible bugs (with make debug).

src/types.h

More customizations

sebastien@ailleret.com home page