ondergetekende / livescrape Goto Github PK

Documentation is an significant part of this project, it needs CI/QA as well. Some ideas:

• Spell checking
• Extract code fragments and check for syntax errors
• Extract code fragments and try to execute them.

Hi Koert,

I've had an unsuccessful stab at allowing one to pass user agent strings to the requests call in livescrape. Any chance you could have a look at this? I'd like to be able to instantiate livescrape with something like

page = ScrapedPage(scrape_url='zzz', headers={'User-agent':'Mozilla something or other'})

Any chance of you having a look at this?

Thanks,

Mat

This is especially useful in conjunction with CssGroup, where the group would find an element, and various selectors could extract attributes from that element.

Livescrape is not the only player in the scraping ecosystem. We will need to explain the reasoning behind its creation, and inform users so they can choose between Livescrape and other projects.

We should at the very least address Scrapy

Hi,

Firstly, thanks for writing this, looks really useful.

I'd like to make a couple of suggestions for the tutorial page.

1. from livescrape import ScrapedPage, Css, CssMulti
   (beginner friendlify it)
2. class project_page = GithubProjectPage(ScrapedPage):
   ^^ your version didn't capitalise Page in GithubProjectPage
3. project_page = GithubProjectPage("ondergetekende", "livesscrape")
   ( double s in livesscrape)
4. print (project_page.contents) returns empty dicts. It would improve the tutorial if something came back here.

I appreciate this isn't as interesting as editing code - I'm happy to send over an error-free tutorial page (or figure out git) if it's easier.

Thanks again,

Mat

I'm assuming that you didn't intend to use a decorator here?

from livescrape import ScrapedPage, Css

class GithubProjectpage(ScrapedPage):
    scrape_url = "https://github.com/%(username)s/%(projectname)s/"
    scrape_args = ("username", "projectname")

    git_repo = @Css("input.input-monospace", attribute="value")

page = GithubProjectpage("python", "cpython")
print(page.description)

For regular attributes, we can use decorator syntax, but that won't work for CssGroup attributes.

Perhaps something like this?

group = CssGroup(".whatever")

@group
@Css("thing")
def foo(self, value, element):
    return None

Whereas the ScrapedPage descendants use attribute access to access the members, CssMulti uses dictionary access. This is especially confusing when the CssMulti exposes a CssLink. In that case you end up with access like article_index.posts[0]['author'].homepage. I feel this should look like article_index.posts[0].author.homepage.

Also, the name CssMulti doesn't really cover it's intended use. I'm considering a syntax like:

class ExampleIndex(ScrapedPage):
    @CssComposite(".article")
    class article(ScrapedSubPage):
        link = CssLink('a.main')

That way the attribute resolving can be made lazy, which will help performance. This method also allows us to use the decorator syntax, and allows definition of non-scraper things.

One disadvantage is that because the class transitions from being a class to an attribute, there is no way to not violate the Pep008 naming scheme. We could work around that by using a two-stage definition scheme, but that feels cumbersome to me.

The CssLink extractor could provide a referer, which may help remote site to identify dead links, and improve the over all credibility of a spider.