Very simple crawler

This commit is contained in:
Mateusz Ogrodowczyk 2022-04-17 21:20:17 +02:00
parent 8883a924b4
commit dd4ffc2faf

110
app/tekstydrugie.hs Normal file
View File

@ -0,0 +1,110 @@
{-# LANGUAGE Arrows, NoMonomorphismRestriction #-}
import ShadowLibrary.Core
import Text.XML.HXT.Core
import Text.XML.HXT.XPath
-- import Text.XML.HXT.Curl
import Data.List
import Data.List.Utils (replace)
import Text.Regex.Posix
import Text.Printf
extractRecords = extractLinksWithText "//section[@id='section-content']//h2/a" -- pary adres-tytuł
>>> second (arr $ replace "\r\n " " ") -- czyścimy drugi element pary, czyli tytuł z niepotrzebnych białych znaków
-- >>> first (arr ((++"tr") . init)) -- modyfikujemy pierwszy element pary, czyli adres URL
>>> first (extractLinksWithText "//div[@class='attachment']/a[contains(@href, '.pdf')]") -- pobieramy stronę z adresu URL i wyciągamy linki z tej strony pasujące do wyrażenia XPathowego
-- ostatecznie wyjdą trójki ((adres URL, tytuł artykułu), tytuł rocznika)
-- ... a tutaj te trójki przerabiamy do docelowej struktury ShadowItem
toShadowItem :: ((String, String), String) -> ShadowItem
toShadowItem ((url, articleTitle), yearlyTitle) =
(defaultShadowItem url title) {
originalDate = Just date,
itype = "periodical",
format = Just "pdf",
finalUrl = url
}
where title = "Teksty Drugie " ++ yearlyTitle ++ " " ++ (replace "\r\n" "" (replace "\r\n " "" articleTitle))
date = getDate url
getDate url =
case url =~~ "/([1-2][0-9][0-9][0-9]|20[0-9][0-9])/" :: Maybe [[String]] of
Just [[_, year]] -> year
otherwise -> error $ "unexpected url: " ++ url
main = do
let start = "http://tekstydrugie.pl/pl/numery-archiwalne/"
let shadowLibrary = ShadowLibrary {logoUrl=Nothing,
lname="Teksty Drugie",
abbrev="TekDru",
lLevel=0,
webpage=start}
let start1 = "http://tekstydrugie.pl/pl/numery-archiwalne/1990/"
let start2 = "http://tekstydrugie.pl/pl/numery-archiwalne/1991/"
let start3 = "http://tekstydrugie.pl/pl/numery-archiwalne/1992/"
let start4 = "http://tekstydrugie.pl/pl/numery-archiwalne/1993/"
let start5 = "http://tekstydrugie.pl/pl/numery-archiwalne/1994/"
let start6 = "http://tekstydrugie.pl/pl/numery-archiwalne/1995/"
let start7 = "http://tekstydrugie.pl/pl/numery-archiwalne/1996/"
let start8 = "http://tekstydrugie.pl/pl/numery-archiwalne/1997/"
let start9 = "http://tekstydrugie.pl/pl/numery-archiwalne/1998/"
let start10 = "http://tekstydrugie.pl/pl/numery-archiwalne/1999/"
let start11 = "http://tekstydrugie.pl/pl/numery-archiwalne/2000/"
let start12 = "http://tekstydrugie.pl/pl/numery-archiwalne/2001/"
let start13 = "http://tekstydrugie.pl/pl/numery-archiwalne/2002/"
let start14 = "http://tekstydrugie.pl/pl/numery-archiwalne/2003/"
let start15 = "http://tekstydrugie.pl/pl/numery-archiwalne/2004/"
let start16 = "http://tekstydrugie.pl/pl/numery-archiwalne/2005/"
let start17 = "http://tekstydrugie.pl/pl/numery-archiwalne/2006/"
let start18 = "http://tekstydrugie.pl/pl/numery-archiwalne/2007/"
let start19 = "http://tekstydrugie.pl/pl/numery-archiwalne/2008/"
let start20 = "http://tekstydrugie.pl/pl/numery-archiwalne/2009/"
let start21 = "http://tekstydrugie.pl/pl/numery-archiwalne/2010/"
let start22 = "http://tekstydrugie.pl/pl/numery-archiwalne/2011/"
let start23 = "http://tekstydrugie.pl/pl/numery-archiwalne/2012/"
let start24 = "http://tekstydrugie.pl/pl/numery-archiwalne/2013/"
let start25 = "http://tekstydrugie.pl/pl/numery-archiwalne/2014/"
let start26 = "http://tekstydrugie.pl/pl/numery-archiwalne/2015/"
let start27 = "http://tekstydrugie.pl/pl/numery-archiwalne/2016/"
let start28 = "http://tekstydrugie.pl/pl/numery-archiwalne/2017/"
let start29 = "http://tekstydrugie.pl/pl/numery-archiwalne/2018/"
let start30 = "http://tekstydrugie.pl/pl/numery-archiwalne/2019/"
let start31 = "http://tekstydrugie.pl/pl/numery-archiwalne/2020/"
let start32 = "http://tekstydrugie.pl/pl/numery-archiwalne/2021/"
extractItemsStartingFromUrl shadowLibrary start1 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start2 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start3 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start4 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start5 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start6 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start7 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start8 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start9 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start10 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start11 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start12 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start13 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start14 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start15 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start16 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start17 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start18 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start19 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start20 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start21 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start22 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start23 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start24 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start25 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start26 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start27 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start28 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start29 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start30 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start31 (extractRecords >>> arr toShadowItem)
extractItemsStartingFromUrl shadowLibrary start32 (extractRecords >>> arr toShadowItem)