frenzy/lib/frenzy/pipeline/scrape_stage.ex

59 lines
1.7 KiB
Elixir

defmodule Frenzy.Pipeline.ScrapeStage do
require Logger
alias Frenzy.Pipeline.Stage
@behaviour Stage
@impl Stage
def apply(_opts, %{url: url} = item_params) do
case get_article_content(url) do
{:ok, content} ->
{:ok, %{item_params | content: content}}
{:error, reason} ->
Logger.warn("Unable to get article content: #{reason}")
item_params
end
end
defp get_article_content(url) when is_binary(url) and url != "" do
Logger.debug("Getting article from #{url}")
url
|> HTTPoison.get()
|> case do
{:ok, response} ->
handle_response(url, response)
{:error, %HTTPoison.Error{reason: reason}} ->
{:error, "HTTPoison error: #{reason}"}
end
end
defp get_article_content(_url), do: {:error, "URL must be a non-empty string"}
defp handle_response(_url, %HTTPoison.Response{status_code: 200, body: body}) do
article = Readability.article(body)
{:ok, Readability.readable_html(article)}
end
defp handle_response(_url, %HTTPoison.Response{status_code: 404}) do
{:error, "404 not found"}
end
defp handle_response(url, %HTTPoison.Response{status_code: status_code, headers: headers})
when status_code in [301, 302] do
{"Location", new_url} = Enum.find(headers, fn {name, _value} -> name == "Location" end)
Logger.debug("Got 301 redirect from #{url} to #{new_url}")
get_article_content(new_url)
end
defp handle_response(_url, %HTTPoison.Response{status_code: 403}) do
{:error, "403 Forbidden"}
end
defp handle_response(_url, %HTTPoison.Response{status_code: status_code} = response) do
{:error, "No handler for response #{inspect(response)}"}
end
end