Fix scraping images w/ URLs w/o schemes

This commit is contained in:
Shadowfacts 2020-02-17 12:09:03 -05:00
parent ecaf4c78b7
commit 4a09ce1cb0
Signed by: shadowfacts
GPG Key ID: 94A5AB95422746E5
3 changed files with 33 additions and 38 deletions

View File

@ -43,4 +43,26 @@ defmodule Frenzy.HTTP do
{:error, reason}
end
end
def resolve_uri(uri, site_uri) when is_binary(site_uri) do
resolve_uri(uri, URI.parse(site_uri))
end
def resolve_uri(%URI{host: nil, path: path}, %URI{} = site_uri) do
%URI{site_uri | path: path}
|> resolve_uri(site_uri)
end
def resolve_uri(%URI{scheme: nil} = uri, %URI{scheme: scheme} = site_uri) do
scheme =
case scheme do
nil -> "https"
_ -> scheme
end
%URI{uri | scheme: scheme}
|> resolve_uri(site_uri)
end
def resolve_uri(uri, _), do: uri
end

View File

@ -112,27 +112,16 @@ defmodule Frenzy.Pipeline.ScrapeStage do
end
end
#
# Generates a helper function for the article with the given URI that takes an HTML element and,
# if it's an <img> element whose src attribute does not have a hostname, adds the hostname and
# scheme to the element.
defp rewrite_image_urls(convert_to_data_uris, %URI{host: host, scheme: scheme}) do
defp rewrite_image_urls(convert_to_data_uris, %URI{host: host, scheme: scheme} = site_uri) do
fn
{"img", attrs} ->
new_attrs =
Enum.map(attrs, fn
{"src", src} ->
case URI.parse(src) do
%URI{host: nil, path: path} ->
new_src =
URI.to_string(%URI{path: path, host: host, scheme: scheme})
|> image_to_data_uri(convert_to_data_uris)
{"src", new_src}
_ ->
{"src", image_to_data_uri(src, convert_to_data_uris)}
end
{"src", image_to_data_uri(src, site_uri, convert_to_data_uris)}
attr ->
attr
@ -148,8 +137,13 @@ defmodule Frenzy.Pipeline.ScrapeStage do
@content_type_allowlist ["image/jpeg", "image/png", "image/heic", "image/heif", "image/tiff"]
# convert images to data URIs so that they're stored by clients as part of the body
defp image_to_data_uri(src, true) do
case HTTP.get(src) do
defp image_to_data_uri(src, site_uri, true) do
absolute_url =
src
|> URI.parse()
|> HTTP.resolve_uri(site_uri)
case HTTP.get(absolute_url) do
{:ok, %HTTPoison.Response{body: body, headers: headers}} ->
{"Content-Type", content_type} =
Enum.find(headers, fn {header, _value} -> header == "Content-Type" end)
@ -165,5 +159,5 @@ defmodule Frenzy.Pipeline.ScrapeStage do
end
end
defp image_to_data_uri(src, false), do: src
defp image_to_data_uri(src, _site_uri, false), do: src
end

View File

@ -105,7 +105,7 @@ defmodule Frenzy.Task.FetchFavicon do
absolute_url =
favicon_url
|> URI.parse()
|> resolve_uri(site_url)
|> HTTP.resolve_uri(URI.parse(site_url))
case HTTP.get(absolute_url) do
{:ok, %HTTPoison.Response{body: body}} ->
@ -118,27 +118,6 @@ defmodule Frenzy.Task.FetchFavicon do
defp fetch_favicon_data(_, _), do: {:error, "No or invalid href for link"}
defp resolve_uri(%URI{host: nil, path: path}, site_url) do
%URI{URI.parse(site_url) | path: path}
|> resolve_uri(site_url)
end
defp resolve_uri(%URI{scheme: nil} = uri, site_url) do
scheme =
case URI.parse(site_url) do
%URI{scheme: scheme} when scheme != "" ->
scheme
_ ->
"https"
end
%URI{uri | scheme: scheme}
|> resolve_uri(site_url)
end
defp resolve_uri(uri, _), do: uri
# from https://github.com/elixir-plug/plug/blob/v1.8.3/lib/plug/request_id.ex#L60
defp generate_task_id() do
binary = <<