Outils_archives/corpus_web
3
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

Feat: pas de balises li et br pour la version CSV

Browse source
This commit is contained in:
cy_altern 2023-05-29 18:43:34 +02:00
parent c9a1a1b36c
commit 6fee512f80
2 changed files with 22 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

17
corpus_web_fonctions.php
View file

@ -330,7 +330,7 @@ function recup_liens_int($html) {
foreach($match as $m) { foreach($match as $m) {
$m[7] = preg_replace('/<(img|(\/)?span)[^>]*>/si', '', $m[7]); $m[7] = preg_replace('/<(img|(\/)?span)[^>]*>/si', '', $m[7]);
// liens exclus // liens exclus
if(in_array($m[3], $url_exclus) || str_contains($m[3], '#comment') || str_contains($m[3], '/user/') || in_array($m[7], $txt_exclus) || str_replace([' ', '&nbsp;'], '', $m[7]) === '') { if(in_array($m[3], $url_exclus) || str_contains($m[3], '#comment') || str_contains($m[3], '/user/') || str_contains($m[3], '/members/') || str_contains($m[3], '/subscription/') || in_array($m[7], $txt_exclus) || str_replace([' ', '&nbsp;'], '', $m[7]) === '') {
continue; continue;
} }
if (!array_key_exists($m[3], $liens_int)) { if (!array_key_exists($m[3], $liens_int)) {
@ -347,3 +347,18 @@ function recup_liens_int($html) {
return count($l_int) ? "<ul><li>".join("</li><li>", $l_int)."</li></ul>" : ''; return count($l_int) ? "<ul><li>".join("</li><li>", $l_int)."</li></ul>" : '';
} }
function sans_li($html) {
$html = str_replace(['<ul>', '</ul>', '<li>'], '', $html);
$html = str_replace('</li>', "\r\n", $html);
$html = str_replace('&nbsp;', " ", $html);
return $html;
}
function sans_br($html) {
$html = str_replace('<br>', "\r\n", $html);
$html = str_replace('&nbsp;', " ", $html);
return $html;
}

12
json_affiche_corpus.json.html
View file

@ -94,12 +94,12 @@
"title": [(#GET{title}|json_encode)], "title": [(#GET{title}|json_encode)],
"desc": [(#GET{description}|json_encode)], "desc": [(#GET{description}|json_encode)],
"keywords": [(#GET{keywords}|json_encode)], "keywords": [(#GET{keywords}|json_encode)],
"trackers": [(#GET{trackers}|json_encode)], "trackers": [(#GET{trackers}|sans_li|json_encode)],
"rezos": [(#GET{rezos}|json_encode)], "rezos": [(#GET{rezos}|sans_li|json_encode)],
"titres": [(#GET{titres}|json_encode)], "titres": [(#GET{titres}|sans_li|json_encode)],
"liens_ext": [(#GET{liens_ext}|json_encode)], "liens_ext": [(#GET{liens_ext}|sans_li|json_encode)],
"liens_int": [(#GET{liens_int}|json_encode)], "liens_int": [(#GET{liens_int}|sans_li|json_encode)],
"text": [(#GET{text}|json_encode)] "text": [(#GET{text}sans_br|json_encode)]
} }
} [(#SET{id_jsonl, #GET{id_jsonl}|plus{1}})] } [(#SET{id_jsonl, #GET{id_jsonl}|plus{1}})]
</BOUCLE_2lignes> </BOUCLE_2lignes>