doesnt work properly

This commit is contained in:
patrycjalazna 2021-04-13 11:02:33 +02:00
parent 4873d78172
commit 931fcdbdb6

View File

@ -13,8 +13,21 @@ import Text.Printf
extractRecords = extractLinksWithText "//p//a[contains(@href,'gov.pl'), text()='View Journal']" -- extractRecords = extractLinks "//a[@class='action' and text() = 'View Journal']"
-- >>> arr ((++"/issue/archive") . init)
-- >>> extractLinksWithText "//div/h4[contains(@class, 'Vol')]/a"
extractRecords = extractLinksWithText "//a[@class='action' and text() = 'View Journal']" -- pary adres-tytuł
>>> second (arr $ replace "\r\n " " ") -- czyścimy drugi element pary, czyli tytuł z niepotrzebnych białych znaków
>>> first (arr ((++"/issue/archive") . init)) -- modyfikujemy pierwszy element pary, czyli adres URL
>>> first (extractLinksWithText "//div/h4[contains(@class, 'Vol')]/a")
-- >>> first (extractLinksWithText "//div[@class='tocTitle']/a")
-- >>> first (arr ((++"/pdf") . init))
-- >>> second (arr $ replace "2020")
-- ostatecznie wyjdą trójki ((adres URL, tytuł artykułu), tytuł rocznika)
toShadowItem :: ((String, String), String) -> ShadowItem toShadowItem :: ((String, String), String) -> ShadowItem
toShadowItem ((url, articleTitle), yearlyTitle) = toShadowItem ((url, articleTitle), yearlyTitle) =
@ -24,19 +37,23 @@ toShadowItem ((url, articleTitle), yearlyTitle) =
format = Just "pdf", format = Just "pdf",
finalUrl = url finalUrl = url
} }
where title = "Geo Journal " ++ yearlyTitle ++ " " ++ (replace "\r\n" "" (replace "\r\n " "" articleTitle)) where title = "Geo Journal OOO" ++ yearlyTitle ++ " AAAAA" ++ (replace "\r\n" "" (replace "\r\n " "" articleTitle))
date = getDate url date = "202020"
-- getDate url =
-- case url =~~ "/(19[0-9][0-9]|20[0-9][0-9])/" :: Maybe [[String]] of
-- Just [[_, year]] -> year
-- otherwise -> error $ "unexpected url: " ++ url
getDate url =
case url =~~ "/(19[0-9][0-9]|20[0-9][0-9])/" :: Maybe [[String]] of
Just [[_, year]] -> year
otherwise -> error $ "unexpected url: " ++ url
main = do main = do
let start = "https://geojournals.pgi.gov.pl" let start = "https://geojournals.pgi.gov.pl"
let shadowLibrary = ShadowLibrary {logoUrl=Nothing, let shadowLibrary = ShadowLibrary {logoUrl=Nothing,
lname="Polish Geological Journals",, lname="Polish Geological Journals",
abbrev="geojour", abbrev="geojour",
lLevel=0, lLevel=0,
webpage=start} webpage=start}