funnelweb 1.0b4

Crawl and parse static sites and import to Plone

FunnelWeb - Content conversion made easy

Easily convert content from existing sites into Plone.

• Code repository: http://github.com/collective/funnelweb

• Questions and comments to http://github.com/collective/funnelweb/issues

• Report bugs at http://github.com/collective/funnelweb/issues

Contents

• FunnelWeb - Content conversion made easy

  • Introduction

  • History

  • Options

    • Crawling - HTML to import

    • Templates

    • Site Analysis

    • Plone Uploading

  • Example: Sphinx to Plone

  • Controlling Logging

  • Working directly with transmogrifier (advanced)

    • Using external blueprints

    • Attributes available in funnelweb pipeline

    • The Funnelweb Pipeline

  • Contributing

• Contributors

• Change history

  • 1.0b5 (2010-12-13)

  • 1.0b4 (2010-12-13)

  • 1.0b3 (2010-11-20)

  • 1.0b2 (2010-11-09)

  • 1.0b1 (2010-11-08)

• Download

Introduction

Funnelweb is very easy to get started with via a few settings in either buildout or the commandline. Funnelweb progresses crawler content through various steps to improve the quality of the final converted site. Each step and it’s configuration options are well thought out and have proved useful on many site conversions over the years.

Funnelweb is also very flexible as it uses a modular transmormation framework underneath which advanced users can use if they they need further steps added to their conversion process.

The simplest way to install is via a buildout recipe (see zc.buildout)

[buildout]
parts += funnelweb

[funnelweb]
recipe = funnelweb
crawler-url=http://www.whitehouse.gov
ploneupload-target=http://admin:admin@localhost:8080/Plone

$> buildout init
$> bin/buildout

The above example will create a script to import content from the whitehouse.gov and upload it to a local Plone site via XML-RPC. This can be run by

$> bin/funnelweb

The work performed by the funnelweb script can be broken down into four sections:

1. Crawling the site including caching locally so subsequent crawls are quicker and filtering out unwanted content

2. Remove boilerplate/templates (automatically or via rules) so just content remains

3. Analysing the site structure to improve the content quality including working out titles, default views, types of objects to create, what to show in navigation etc

4. Uploading to the CMS such as Plone, or saveing cleaned HTML to local directory

History

• 2008 Built to import large corporate intranet

• 2009 released pretaweb.funnelweb (deprecated). Built into Plone UI > Actions > Import

• 2010 Split blueprints into transmogrify.* release on pypi

• 2010 collective.developermanual sphinx to Plone uses funnelweb blueprints

• 2010 funnelweb Recipe + Script released

Options

Funnelweb is organised as a series of steps through which crawled items pass before eventually being uploaded. Each step as one or more configuration options so you can customise import process for your needs. Almost all imports will require some level of configurations.

The first part of each configuration key is the step e.g. crawler. The second part is the particular configuration option for that particular step. e.g. url. This is then followed by = and value or values.

The configuration options can either be given as part of the buildout part e.g.

[buildout]
parts += funnelweb

[funnelweb]
recipe = funnelweb
crawler-url=http://www.whitehouse.gov

or the same option can be overridden via the command line

$> bin/funnelweb --crawler:url=http://www.whitehouse.gov

The full list of steps that can be configured is

1. Crawling

• crawler

• cache

• typeguess

• drop

2. Templates

• template1

• template2

• template3

• template4

• templateauto

3. Site Analysis

• indexguess

• titleguess

• attachmentguess

• urltidy

• addfolders

• changetype

4. Uploading

• ploneupload

• ploneupdate

• plonehide

• publish

• plonepublish

• plonealias

• ploneprune

• localupload

The most common configuration options for these steps are detailed below.

Crawling - HTML to import

Funnelweb imports HTML either from a live website, from a folder on disk, or a folder on disk with HTML which was retrieved from a live website and may still have absolute links refering to that website.

Funnelweb can only import things it can crawl, i.e. content that is linked from HTML. If your site contains javascript links or password protected content, then you may have to perform some extra steps to get funnelweb to crawl your content.

To crawl a live website, supply the crawler with a base HTTP URL to start crawling from. This URL must be the URL which all the other URLs you want from the site start with.

For example

$> bin/funnelweb --crawler:url=http://www.whitehouse.gov --crawler:maxsize=50  --ploneupload=http://admin:admin@localhost:8080/Plone

will restrict the crawler to the first 50 pages and then convert the content into a local Plone site.

The site you crawl will be cached locally, so if you run funnelweb again it will run much quicker. If you’d like to disable the local caching use

$> bin/funnelweb --cache:output=

If you’d like to reset the cache, refreshing it’s data, set the crawlers cache to nothing

$> bin/funnelweb --crawler:cache=

By default the cache is stored in var/funnelwebcache/{site url}/. You can set this to another directory using:

$> bin/funnelweb --cache:output=my_new_dir

You can also crawl a local directory of HTML with relative links by just using a file:// style URL

$> bin/funnelweb --crawler:url=file:///mydirectory

or if the local directory contains HTML saved from a website and might have absolute URLs in it, the you can set this as the cache. The crawler will always look up the cache first

$> bin/funnelweb --crawler:url=http://therealsite.com --crawler:cache=mydirectory

The following will not crawl anything larger than 4Mb

$> bin/funnelweb --crawler:max=400000

To skip crawling links by regular expression

[funnelweb]
recipe = funnelweb
crawler-url=http://www.whitehouse.gov
crawler-ignore = \.mp3
                 \.mp4

If funnelweb is having trouble parsing the HTML of some pages, you can preprocesses the HTML before it is parsed. e.g.

[funnelweb]
recipe = funnelweb
crawler-patterns = (<script>)[^<]*(</script>)
crawler-subs = \1\2

If you’d like to skip processing links with certain mimetypes you can use the drop:condition option. This TALES expression determines what will be processed further

[funnelweb]
recipe = funnelweb
drop-condition: python:item.get('_mimetype') not in ['application/x-javascript','text/css','text/plain','application/x-java-byte-code'] and item.get('_path','').split('.')[-1] not in ['class']

Templates

Funnelweb has a built-in clustering algorithm that tries to automatically extract the content from the HTML template. This is slow and not always effective. Often you will need to input your own template extraction rules.

If you’d like to turn off the automatic templates

$> bin/funnelweb --templateauto:condition=python:False

Rules are in the form of

(title|description|text|anything) = (text|html|optional) XPath

For example

[funnelweb]
recipe = funnelweb
crawler-site_url=http://www.whitehouse.gov
ploneupload-target=http://admin:admin@localhost:8080/Plone
template1-title       = text //div[@class='body']//h1[1]
template1-_delete1    = optional //div[@class='body']//a[@class='headerlink']
template1-_delete2    = optional //div[contains(@class,'admonition-description')]
template1-description = text //div[contains(@class,'admonition-description')]//p[@class='last']
template1-text        = html //div[@class='body']

Note that for a single template e.g. template1, ALL of the XPaths need to match otherwise that template will be skipped and the next template tried. If you’d like to make it so that a single XPath isn’t nessary for the template to match then use the keyword optional instead of text or html before the XPath.

In the default pipeline there are four templates called template1, template2, template3 and template4.

When an XPath is applied within a single template, the HTML it matches will be removed from the page. Another rule in that same template can’t match the same HTML fragment.

If a content part is not useful to Plone (e.g. redundant text, title or description) it is a way to effectively remove that HTML from the content.

To help debug your template rules you can set debug mode

$> bin/funnelweb --template1:debug --template2:debug

Setting debug mode on templateauto will give you details about the rules it uses.

$> bin/funnelweb --templateauto:debug
...
DEBUG:templateauto:'icft.html' discovered rules by clustering on 'http://...'
Rules:
      text= html //div[@id = "dal_content"]//div[@class = "content"]//p
      title= text //div[@id = "dal_content"]//div[@class = "content"]//h3
Text:
      TITLE: ...
      MAIN-10: ...
      MAIN-10: ...
      MAIN-10: ...

For more information about XPath see

• http://www.w3schools.com/xpath/default.asp

• http://blog.browsermob.com/2009/04/test-your-selenium-xpath-easily-with-firebug/

Site Analysis

In order to provide a cleaner-looking Plone site, there are several options to analyse the entire crawled site and clean it up. These are turned off by default.

To determine if an item is a default page for a container (it has many links to items in that container, even if not contained in that folder), and then move it to that folder, use

$> bin/funnelweb --indexguess:condition=python:True

You can automatically find better page titles by analysing backlink text

[funnelweb]
recipe = funnelweb
titleguess-condition = python:True
titleguess-ignore =
      click
      read more
      close
      Close
      http:
      https:
      file:
      img

The following will find items only referenced by one page and move them into a new folder with the page as the default view.

$> bin/funnelweb --attachmentguess:condition=python:True

or the following will only move attachments that are images and use index-html as the new name for the default page of the newly created folder

[funnelweb]
recipe = funnelweb
attachmentguess-condition = python: subitem.get('_type') in ['Image']
attachmentguess-defaultpage = index-html

The following will tidy up the URLs based on a TALES expression

$> bin/funnelweb --urltidy:link_expr="python:item['_path'].endswith('.html') and item['_path'][:-5] or item['_path']"

If you’d like to move content around before it’s uploaded you can use the urltidy step as well e.g.

$> bin/funnelweb --urltidy:link_expr=python:item['_path'].startswith('/news') and '/otn/news'+item['path'][5:] or item['_path']

Plone Uploading

Uploading happens via remote XML-RPC calls so can be done to a live running site anywhere.

To set where a the site will be uploaded to use

$> bin/funnelweb --ploneupload:target=http://username:[e-mail]/myfolder

Currently only basic authentication via setting the username and password in the url is supported. If no target is set then the site will be crawled but not uploaded.

If you’d like to change the type of what’s uploaded

$> bin/funnelweb --changetype:value=python:{'Folder':'HelpCenterReferenceManualSection','Document':HelpCenterLeafPage}.get(item['_type'],item['_type'])

This will set a new value for the type of the item. You could make this conditional e.g

$> bin/funnelweb --changetype:condition=python:item['_path].startswith('/news')

or by using a more complex expression for the new type

$> bin/funnelweb –changetype:value=python:item[‘_path’].startswith(‘/news’) and ‘NewNewsType’ or item[‘_type]

By default, funnelweb will automatically create Plone aliases based on the original crawled URLs, so that any old links will automatically be redirected to the new cleaned-up urls. You can disable this by

$> bin/funnelweb --plonealias:target=

You can change what items get published to which state by setting the following

[funnelweb]
recipe = funnelweb
publish-value = python:["publish"]
publish-condition = python:item.get('_type') != 'Image' and not options.get('disabled')

Funnelweb will hide certain items from Plone’s navigation if that item was only ever linked to from within the content area. You can disable this behavior by

$> bin/funnelweb --plonehide:target=

You can get a local file representation of what will be uploaded by using the following

$> bin/funnelweb --localupload:output=var/mylocaldir

Example: Sphinx to Plone

As an example the following buildout recipe will create a funnelweb script that will convert a regular sphix documentation into remote plone content inside a PloneHelpCenter

[buildout]
parts += sphnix funnelweb

[sphinx]
recipe = collective.recipe.sphinxbuilder
#doc-directory = .
outputs = html
source = ${buildout:directory}/source
build = ${buildout:directory}/build
eggs =
  Sphinx
  Docutils
  roman
  Pygments


[funnelweb]
recipe = funnelweb
crawler-url=file://${buildout:directory}/build/html
crawler-ignore=
        cgi-bin
        javascript:
        _static
        _sources
        genindex\.html
        search\.html
        saesrchindex\.js
cache-output =

# Permalinks are Sphinx's on-hover makers that are not used in Plone. We have this dummy rule here to parse
# them out from body so that they do not corrupt "text" field

template1-title = text //div[@class='body']//h1[1]
template1-_permalink = text //div[@class='body']//a[@class='headerlink']
template1-text = html //div[@class='body']
template1-description = optional //div[contains(@class,'admonition-description')]//p[@class='last']
                     //div[contains(@class,'admonition-description')]
templateauto-condition = python:False

attachmentguess-condition = python: subitem.get('_type') in ['Image']
attachmentguess-defaultpage = index

changetype-value=python:{'Folder':'HelpCenterReferenceManualSection','Document':'HelpCenterLeafPage'}.get(item['_type'],item['_type'])

localupload-output=${buildout:directory}/ploneout

# All folderish content should be checked if they contain
# any items on the remote site which are not presented locally. including base folder
ploneprune-condition=python:item.get('_type') in ['HelpCenterReferenceManualSection','HelpCenterReferenceManual'] or item['_path'] == ''

ploneupload-target=http://admin:admin@localhost/Plone

Controlling Logging

You can show additional debug output on any particular set by setting a debug commandline switch. For instance to see see additional details about template matching failures

$> bin/funnelweb --template1:debug

Working directly with transmogrifier (advanced)

You might need to insert further transformation steps for your particular conversion usecase. To do this, you can extend funnelweb’s underlying transmogrifier pipeline. Funnelweb uses a transmogrifier pipeline to perform the needed transformations and all commandline and recipe options refer to options in the pipeline.

You can view pipeline and all its options via the following command

$> bin/funnelweb --pipeline

You can also save this pipeline and customise it for your own needs

$> bin/funnelweb --pipeline > pipeline.cfg
$> {edit} pipeline.cfg
$> bin/funnelweb --pipeline=pipeline.cfg

Customising the pipeline allows you add your own personal transformations which haven’t been pre-considered by the standard funnelweb tool.

See transmogrifier documentation to see how to add your own blueprints or add blueprints that already exist to your custom pipeline.

Using external blueprints

If you have decided you need to customise your pipeline and you want to install transformation steps that use blueprints not already included in funnelweb or transmogrifier, you can include them using the eggs option in a funnelweb buildout part

[funnelweb]
recipe = funnelweb
eggs = myblueprintpackage
pipeline = mypipeline.cfg

However, this only works if your blueprint package includes the following setuptools entrypoint in its setup.py

entry_points="""
      [z3c.autoinclude.plugin]
      target = transmogrify
      """,
      )

Note

Some transmogrifier blueprints assume they are running inside a Plone process such as those in plone.app.transmogrifier (see http://pypi.python.org/pypi/plone.app.transmogrifier). Funnelweb doesn’t run inside a Plone process so these blueprints won’t work. If you want upload content into Plone, you can instead use transmogrify.ploneremote which provides alternative implementations which will upload content remotely via XML-RPC. transmogrify.ploneremote is already included in funnelweb as it is what funnelweb’s default pipeline uses.

Attributes available in funnelweb pipeline

When using the default blueprints in funnelweb the following are some of the attributes that will become attached to the items that each blueprint has access to. These can be used in the various condition statements etc. as well as your own blueprints.

_site_url

The base of the url as passed into the webcrawler

_path

The remainder of the URL. _site_url + _path = URL

_mimetype

The mimetype as returned by the crawler

_content

The content of the item crawled, include image, file or HTML data.

_orig_path

The original path of the item that was crawled. This is useful for setting redirects so you don’t get 404 errors after migrating content.

_sort_order

An integer representing the order in which this item was crawled. Helps to determine what order items should be sorted in folders created on the server if your site has navigation which has links ordered top to bottom.

_type

The type of object to be created as returned by the “typeguess” step

title, description, text, etc.

The template steps will typically create fields with content in them taken from _content

_template

The template steps will leave the HTML that wasn’t seperated out into different fields in this attribute.

_defaultpage

Set on an Folder item where you want to tell the uploading steps to set the containing item mentioned in _defaultpage to be the default page shown on that folder instead of a content listing.

_transitions

Specify the workflow action you’d like to make on an item after it’s uploaded or updated.

_origin

This is used internally with the transmogrify.siteanalysis.relinker blueprint as a way to tell it that you have changed the _path and you now want the relinker to find any links that refer to _origin to now point to _path.

The Funnelweb Pipeline

see http://github.com/collective/funnelweb/blob/master/funnelweb/runner/pipeline.cfg or type

$> bin/funnelweb --pipeline

Contributing

• Code repository: http://github.com/collective/funnelweb

• Questions and comments to http://github.com/collective/funnelweb/issues

• Report bugs at http://github.com/collective/funnelweb/issues

The code of funnelweb itself is fairly minimal. It just sets up and runs a transmogrifier pipeline. The hard work is actually done by five packages which each contain one or more transmogrifier blueprints. These are:

Webcrawler

http://pypi.python.org/pypi/transmogrify.webcrawler https://github.com/djay/transmogrify.webcrawler

HTMLContentExtractor

http://pypi.python.org/pypi/transmogrify.htmlcontentextractor https://github.com/djay/transmogrify.htmlcontentextractor

SiteAnalyser

http://pypi.python.org/pypi/transmogrify.siteanalyser https://github.com/djay/transmogrify.siteanalyser

PathSorter

http://pypi.python.org/pypi/transmogrify.pathsorter https://github.com/djay/transmogrify.pathsorter

PloneRemote

http://pypi.python.org/pypi/transmogrify.ploneremote https://github.com/djay/transmogrify.ploneremote

Each has it’s own issue tracker and I will accept pull requests for new functionality or bug fixes. The current state of documentation and testing is not yet at a high level.

Contributors

“Dylan Jay”, Author “Vitaliy Podoba”, Contributor “Rok Garbas”, Contributor “Mikko Ohtamaa”, Contributor “Tim Knapp”, Contributor

Change history

1.0b5 (2010-12-13)

• fix ordering of commandline help

• fix help for @debug

1.0b4 (2010-12-13)

• fix encoding problems caused by cache

• better debugging

• commandline to turn on debug info

• script install uses buildout part name

• extra documentation

• commandline help

1.0b3 (2010-11-20)

• fixed –pipeline option

• fixed eggs= options

• added prune support

• removed transmogrify.htmltesting as a dependency

• improved documentation [Jean Jordaan]

• moved main repository to github collective https://github.com/collective/funnelweb

1.0b2 (2010-11-09)

• Removed z3c.recipe.scripts as a dependency since it creates version conflicts with older zope installs. [“Dylan Jay”]

• Make default cache be put in domain specific directory [“Dylan Jay”]

• Put conditions on site analyser and turn off by default

1.0b1 (2010-11-08)

• Initial release tying togeather new and original funnelweb recipes into documented commandline/buildout interface [“Dylan Jay”]

Download