corpus_web/json_affiche_corpus.json.html
2023-05-29 19:05:25 +02:00

107 lines
4.4 KiB
HTML

#HTTP_HEADER{Content-Type: application/json; charset=#CHARSET}
[(#ENV{id}|setenv{id_jsonl})]
[{
"header":{
"id": "id jsonl",
"site": "Site",
"date_aspi": "Date archive",
"num_page": "Numéro page",
"title": "Meta title",
"desc": "Meta description",
"keywords": "Meta keywords",
"trackers": "Trackers",
"rezos": "Rézos socios",
"titres": "titres",
"liens_ext": "Liens externes",
"liens_int": "Liens internes",
"text": "Texte"
},
"filtreCol" : {
"site" : "select",
"date_aspi": "select",
"num_page": "select",
"title": "input",
"desc": "input",
"keywords": "input",
"trackers": "select",
"rezos": "select",
"titres": "input",
"liens_ext": "input",
"liens_int": "input",
"text": "input"
},
"classes":{
"num_page": "w80p",
"title": "w120p",
"desc": "w120p",
"keywords": "w200p",
"trackers": "w100p",
"rezos": "w100p",
"liens_int": "w400p maxw400p",
"text": "minw400p"
}
} #SET{id_jsonl,1}
<BOUCLE_corpus(DATA){source ls, #CHEMIN{tmp}/#CONFIG{corpus_web/repertoire_corpus,corpus_jsonl}/*}{par basename}>
[(#SET{site,#FILE|basename|recup_site})] [(#SET{date_aspi,#FILE|basename|recup_aspi})]
#SET{liste_lignes,#LISTE{1}} #SET{trop_gros,#CONFIG{corpus_web/taille_max}|mult{1000000}}
#SET{nb_lignes, #FILE|nombre_ligne} [(#GET{nb_lignes}|!={1}|oui) #SET{liste_lignes,#LISTE{1,#GET{nb_lignes}}}]
<BOUCLE_2lignes(DATA){enum 1, #GET{nb_lignes}, #CONFIG{corpus_web/pas_corpus, 10}}{si #SIZE|<{#GET{trop_gros}}}>
[(#SET{num_ligne,[(#COMPTEUR_BOUCLE|=={1}|?{1,#GET{nb_lignes}})]})]
[(#SET{content,[(#FILE|affiche_ligne{#GET{num_ligne}}|json_decode{true})]})]
#SET{num_page,Page #VALEUR} #SET{title,''} #SET{keywords,''} #SET{description,''} #SET{titres,''} #SET{trackers,''} #SET{rezos,''} #SET{liens_ext,''} #SET{liens_int,''}
<BOUCLE_content(DATA){source table, #GET{content}}>
[(#CLE|=={extractionContent}|oui)
<BOUCLE_ss(DATA){source table, #VALEUR}>
[(#CLE|=={boilerpipe:text}|oui)
[(#SET{text, [(#VALEUR|print|supprimer_com|replace{<br>,§§}|replace{<br />,§§}|replace{<,&lt;}|replace{>,&gt;}|replace{§§,<br>}|replace{"\n", <br>})]})]
]
[(#CLE|=={htmlmeta:head:title}|oui) [(#SET{title, [(#VALEUR|print|replace{<br>,§§}|replace{<br />,§§}|replace{<,&lt;}|replace{>,&gt;}|replace{§§,<br>}|replace{"\n", <br>})]})] ]
[(#CLE|=={htmlmeta:head:keywords}|oui) [(#SET{keywords, [(#VALEUR|print|replace{<br>,§§}|replace{<br />,§§}|replace{<,&lt;}|replace{>,&gt;}|replace{§§,<br>}|replace{"\n", <br>}|replace{',', ', '})]})] ]
[(#CLE|=={htmlmeta:head:description}|oui) [(#SET{description, [(#VALEUR|print|replace{<br>,§§}|replace{<br />,§§}|replace{<,&lt;}|replace{>,&gt;}|replace{§§,<br>}|replace{"\n", <br>})]})] ]
</BOUCLE_ss>
][(#CLE|=={htmlBytes}|oui)
[(#SET{titres, [(#VALEUR|base64_decode|recup_titres)]})]
[(#SET{trackers, [(#VALEUR|base64_decode|trouve_trackers)]})]
[(#SET{rezos, [(#VALEUR|base64_decode|trouve_rezos)]})]
[(#SET{liens_ext, [(#VALEUR|base64_decode|recup_liens_ext)]})]
[(#SET{liens_int, [(#VALEUR|base64_decode|recup_liens_int)]})]
]
</BOUCLE_content>
,{
"html": {
"id": [(#GET{id_jsonl})],
"site" : [(#GET{site}|json_encode)],
"date_aspi": [(#GET{date_aspi}|json_encode)],
"num_page": [(#GET{num_page}|json_encode)],
"title": [(#GET{title}|json_encode)],
"desc": [(#GET{description}|json_encode)],
"keywords": [(#GET{keywords}|json_encode)],
"trackers": [(#GET{trackers}|json_encode)],
"rezos": [(#GET{rezos}|json_encode)],
"titres": [(#GET{titres}|json_encode)],
"liens_ext": [(#GET{liens_ext}|json_encode)],
"liens_int": [(#GET{liens_int}|json_encode)],
"text": [(#GET{text}|json_encode)]
},
"classes": {
"titre":"transcription"
},
"search": {
"id": [(#GET{id_jsonl})],
"site" : [(#GET{site}|json_encode)],
"date_aspi": [(#GET{date_aspi}|json_encode)],
"num_page": [(#GET{num_page}|json_encode)],
"title": [(#GET{title}|json_encode)],
"desc": [(#GET{description}|json_encode)],
"keywords": [(#GET{keywords}|json_encode)],
"trackers": [(#GET{trackers}|sans_li|json_encode)],
"rezos": [(#GET{rezos}|sans_li|json_encode)],
"titres": [(#GET{titres}|sans_li|json_encode)],
"liens_ext": [(#GET{liens_ext}|sans_li|json_encode)],
"liens_int": [(#GET{liens_int}|sans_li|json_encode)],
"text": [(#GET{text}|sans_br|json_encode)]
}
} [(#SET{id_jsonl, #GET{id_jsonl}|plus{1}})]
</BOUCLE_2lignes>
</BOUCLE_corpus>
]