Cascadia

Inspired by, and mostly a direct translation of, the Cascadia CSS Selector library, written in Go, by @andybalhom.

This package depends on the Gumbo.jl package by @porterjamesj, which is a Julia wrapper around Google's Gumbo HTML parser library

Usage is simple. Use Gumbo to parse an HTML string into a document, create a Selector from a string, and then use eachmatch to get the nodes in the document that match the selector. Alternatively, use sel"<selector string>" to do the same thing as Selector. The eachmatch function returns an array of elements which match the selector. If no match is found, a zero element array is returned. For unique matches, the array contains one element. Thus, check the length of the array to test whether a selector matches.

using Cascadia
using Gumbo

n=parsehtml("<p id=\"foo\"><p id=\"bar\">")
s=Selector("#foo")
sm = sel"#foo"
eachmatch(s, n.root)
# 1-element Array{Gumbo.HTMLNode,1}:
#  Gumbo.HTMLElement{:p}

eachmatch(sm, n.root)
# 1-element Array{Gumbo.HTMLNode,1}:
#  Gumbo.HTMLElement{:p}

Note: The top level matching function name has changed from matchall in v0.6 to eachmatch in v0.7 and higher to reflect the change in Julia base.

The primary use case for this library is to enable webscraping -- the automatic extraction of information from html pages. As an example, consider the following code, which returns a list of questions that have been tagged with julia-lang on StackOverflow.

using Cascadia, Gumbo, HTTP

r = HTTP.get("http://stackoverflow.com/questions/tagged/julia-lang")
h = parsehtml(String(r.body))

qs = eachmatch(Selector(".question-summary"),h.root)

println("StackOverflow Julia Questions (votes  answered?  url)")

for q in qs
    votes = nodeText(eachmatch(Selector(".votes .vote-count-post "), q)[1])
    answered = length(eachmatch(Selector(".status.answered"), q)) > 0 || length(eachmatch(Selector(".status.answered-accepted"), q)) > 0
    href = eachmatch(Selector(".question-hyperlink"), q)[1].attributes["href"]
    println("$votes  $answered  http://stackoverflow.com$href")
end

This code produces the following output:

StackOverflow Julia Questions (votes  answered?  url)

0  false  http://stackoverflow.com/questions/59361325/how-to-get-a-rolling-window-regression-in-julia
0  true  http://stackoverflow.com/questions/59356818/how-i-translate-python-code-into-julia-code
-2  false  http://stackoverflow.com/questions/59354720/how-to-fix-this-error-in-julia-throws-same-error-for-all-packages-not-found-i
-1  true  http://stackoverflow.com/questions/59354407/julia-package-for-geocoding
1  false  http://stackoverflow.com/questions/59350631/jupyter-lab-precompile-error-for-kernel-1-0-after-adding-kernel-1-3
0  true  http://stackoverflow.com/questions/59348461/genie-framework-does-not-install-under-julia-1-2
...
2  true  http://stackoverflow.com/questions/59300202/julia-package-install-fail-with-please-specify-by-known-name-uuid
2  false  http://stackoverflow.com/questions/59297379/how-do-i-transfer-my-packages-after-installing-a-new-julia-version

Note that this returns the elements on the first page of the query results. Getting the values from subsequent pages is left as an exercise for the reader.

This library should work with almost all CSS selectors. Please raise an issue if you find any that don't work. However, note that CSS pseudo elements are not yet supported.

Specifically, the following selector types are tested, and known to work.