mirror of https://gitlab.com/bashrc2/epicyon
998 lines
40 KiB
Python
998 lines
40 KiB
Python
__filename__ = "webapp_search.py"
|
|
__author__ = "Bob Mottram"
|
|
__license__ = "AGPL3+"
|
|
__version__ = "1.2.0"
|
|
__maintainer__ = "Bob Mottram"
|
|
__email__ = "bob@libreserver.org"
|
|
__status__ = "Production"
|
|
__module_group__ = "Web Interface"
|
|
|
|
import os
|
|
from shutil import copyfile
|
|
import urllib.parse
|
|
from datetime import datetime
|
|
from utils import getBaseContentFromPost
|
|
from utils import isAccountDir
|
|
from utils import getConfigParam
|
|
from utils import getFullDomain
|
|
from utils import isEditor
|
|
from utils import loadJson
|
|
from utils import getDomainFromActor
|
|
from utils import getNicknameFromActor
|
|
from utils import locatePost
|
|
from utils import isPublicPost
|
|
from utils import firstParagraphFromString
|
|
from utils import searchBoxPosts
|
|
from utils import getAltPath
|
|
from utils import acctDir
|
|
from utils import localActorUrl
|
|
from skills import noOfActorSkills
|
|
from skills import getSkillsFromList
|
|
from categories import getHashtagCategory
|
|
from feeds import rss2TagHeader
|
|
from feeds import rss2TagFooter
|
|
from webapp_utils import setCustomBackground
|
|
from webapp_utils import htmlKeyboardNavigation
|
|
from webapp_utils import htmlHeaderWithExternalStyle
|
|
from webapp_utils import htmlFooter
|
|
from webapp_utils import getSearchBannerFile
|
|
from webapp_utils import htmlPostSeparator
|
|
from webapp_utils import htmlSearchResultShare
|
|
from webapp_post import individualPostAsHtml
|
|
from webapp_hashtagswarm import htmlHashTagSwarm
|
|
|
|
|
|
def htmlSearchEmoji(cssCache: {}, translate: {},
|
|
baseDir: str, httpPrefix: str,
|
|
searchStr: str) -> str:
|
|
"""Search results for emoji
|
|
"""
|
|
# emoji.json is generated so that it can be customized and the changes
|
|
# will be retained even if default_emoji.json is subsequently updated
|
|
if not os.path.isfile(baseDir + '/emoji/emoji.json'):
|
|
copyfile(baseDir + '/emoji/default_emoji.json',
|
|
baseDir + '/emoji/emoji.json')
|
|
|
|
searchStr = searchStr.lower().replace(':', '').strip('\n').strip('\r')
|
|
cssFilename = baseDir + '/epicyon-profile.css'
|
|
if os.path.isfile(baseDir + '/epicyon.css'):
|
|
cssFilename = baseDir + '/epicyon.css'
|
|
|
|
emojiLookupFilename = baseDir + '/emoji/emoji.json'
|
|
|
|
# create header
|
|
instanceTitle = \
|
|
getConfigParam(baseDir, 'instanceTitle')
|
|
emojiForm = htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
emojiForm += '<center><h1>' + \
|
|
translate['Emoji Search'] + \
|
|
'</h1></center>'
|
|
|
|
# does the lookup file exist?
|
|
if not os.path.isfile(emojiLookupFilename):
|
|
emojiForm += '<center><h5>' + \
|
|
translate['No results'] + '</h5></center>'
|
|
emojiForm += htmlFooter()
|
|
return emojiForm
|
|
|
|
emojiJson = loadJson(emojiLookupFilename)
|
|
if emojiJson:
|
|
results = {}
|
|
for emojiName, filename in emojiJson.items():
|
|
if searchStr in emojiName:
|
|
results[emojiName] = filename + '.png'
|
|
for emojiName, filename in emojiJson.items():
|
|
if emojiName in searchStr:
|
|
results[emojiName] = filename + '.png'
|
|
headingShown = False
|
|
emojiForm += '<center>'
|
|
msgStr1 = translate['Copy the text then paste it into your post']
|
|
msgStr2 = ':<img loading="lazy" class="searchEmoji" src="/emoji/'
|
|
for emojiName, filename in results.items():
|
|
if os.path.isfile(baseDir + '/emoji/' + filename):
|
|
if not headingShown:
|
|
emojiForm += \
|
|
'<center><h5>' + msgStr1 + \
|
|
'</h5></center>'
|
|
headingShown = True
|
|
emojiForm += \
|
|
'<h3>:' + emojiName + msgStr2 + \
|
|
filename + '"/></h3>'
|
|
emojiForm += '</center>'
|
|
|
|
emojiForm += htmlFooter()
|
|
return emojiForm
|
|
|
|
|
|
def _matchSharedItem(searchStrLowerList: [],
|
|
sharedItem: {}) -> bool:
|
|
"""Returns true if the shared item matches search criteria
|
|
"""
|
|
for searchSubstr in searchStrLowerList:
|
|
searchSubstr = searchSubstr.strip()
|
|
if sharedItem.get('location'):
|
|
if searchSubstr in sharedItem['location'].lower():
|
|
return True
|
|
if searchSubstr in sharedItem['summary'].lower():
|
|
return True
|
|
elif searchSubstr in sharedItem['displayName'].lower():
|
|
return True
|
|
elif searchSubstr in sharedItem['category'].lower():
|
|
return True
|
|
return False
|
|
|
|
|
|
def _htmlSearchResultSharePage(actor: str, domainFull: str,
|
|
callingDomain: str, pageNumber: int,
|
|
searchStrLower: str, translate: {},
|
|
previous: bool) -> str:
|
|
"""Returns the html for the previous button on shared items search results
|
|
"""
|
|
postActor = getAltPath(actor, domainFull, callingDomain)
|
|
# previous page link, needs to be a POST
|
|
if previous:
|
|
pageNumber -= 1
|
|
titleStr = translate['Page up']
|
|
imageUrl = 'pageup.png'
|
|
else:
|
|
pageNumber += 1
|
|
titleStr = translate['Page down']
|
|
imageUrl = 'pagedown.png'
|
|
sharedItemsForm = \
|
|
'<form method="POST" action="' + postActor + '/searchhandle?page=' + \
|
|
str(pageNumber) + '">\n'
|
|
sharedItemsForm += \
|
|
' <input type="hidden" ' + 'name="actor" value="' + actor + '">\n'
|
|
sharedItemsForm += \
|
|
' <input type="hidden" ' + 'name="searchtext" value="' + \
|
|
searchStrLower + '"><br>\n'
|
|
sharedItemsForm += \
|
|
' <center>\n' + ' <a href="' + actor + \
|
|
'" type="submit" name="submitSearch">\n'
|
|
sharedItemsForm += \
|
|
' <img loading="lazy" ' + 'class="pageicon" src="/icons' + \
|
|
'/' + imageUrl + '" title="' + titleStr + \
|
|
'" alt="' + titleStr + '"/></a>\n'
|
|
sharedItemsForm += ' </center>\n'
|
|
sharedItemsForm += '</form>\n'
|
|
return sharedItemsForm
|
|
|
|
|
|
def _htmlSharesResult(baseDir: str,
|
|
sharesJson: {}, pageNumber: int, resultsPerPage: int,
|
|
searchStrLowerList: [], currPage: int, ctr: int,
|
|
callingDomain: str, httpPrefix: str, domainFull: str,
|
|
contactNickname: str, actor: str,
|
|
resultsExist: bool, searchStrLower: str, translate: {},
|
|
sharesFileType: str) -> (bool, int, int, str):
|
|
"""Result for shared items search
|
|
"""
|
|
sharedItemsForm = ''
|
|
if currPage > pageNumber:
|
|
return resultsExist, currPage, ctr, sharedItemsForm
|
|
|
|
for name, sharedItem in sharesJson.items():
|
|
if _matchSharedItem(searchStrLowerList, sharedItem):
|
|
if currPage == pageNumber:
|
|
# show individual search result
|
|
sharedItemsForm += \
|
|
htmlSearchResultShare(baseDir, sharedItem, translate,
|
|
httpPrefix, domainFull,
|
|
contactNickname,
|
|
name, actor, sharesFileType,
|
|
sharedItem['category'])
|
|
if not resultsExist and currPage > 1:
|
|
# show the previous page button
|
|
sharedItemsForm += \
|
|
_htmlSearchResultSharePage(actor, domainFull,
|
|
callingDomain,
|
|
pageNumber,
|
|
searchStrLower,
|
|
translate, True)
|
|
resultsExist = True
|
|
ctr += 1
|
|
if ctr >= resultsPerPage:
|
|
currPage += 1
|
|
if currPage > pageNumber:
|
|
# show the next page button
|
|
sharedItemsForm += \
|
|
_htmlSearchResultSharePage(actor, domainFull,
|
|
callingDomain,
|
|
pageNumber,
|
|
searchStrLower,
|
|
translate, False)
|
|
return resultsExist, currPage, ctr, sharedItemsForm
|
|
ctr = 0
|
|
return resultsExist, currPage, ctr, sharedItemsForm
|
|
|
|
|
|
def htmlSearchSharedItems(cssCache: {}, translate: {},
|
|
baseDir: str, searchStr: str,
|
|
pageNumber: int,
|
|
resultsPerPage: int,
|
|
httpPrefix: str,
|
|
domainFull: str, actor: str,
|
|
callingDomain: str,
|
|
sharedItemsFederatedDomains: [],
|
|
sharesFileType: str) -> str:
|
|
"""Search results for shared items
|
|
"""
|
|
currPage = 1
|
|
ctr = 0
|
|
sharedItemsForm = ''
|
|
searchStrLower = urllib.parse.unquote(searchStr)
|
|
searchStrLower = searchStrLower.lower().strip('\n').strip('\r')
|
|
searchStrLowerList = searchStrLower.split('+')
|
|
cssFilename = baseDir + '/epicyon-profile.css'
|
|
if os.path.isfile(baseDir + '/epicyon.css'):
|
|
cssFilename = baseDir + '/epicyon.css'
|
|
|
|
instanceTitle = \
|
|
getConfigParam(baseDir, 'instanceTitle')
|
|
sharedItemsForm = \
|
|
htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
if sharesFileType == 'shares':
|
|
titleStr = translate['Shared Items Search']
|
|
else:
|
|
titleStr = translate['Wanted Items Search']
|
|
sharedItemsForm += \
|
|
'<center><h1>' + \
|
|
'<a href="' + actor + '/search">' + titleStr + '</a></h1></center>'
|
|
resultsExist = False
|
|
for subdir, dirs, files in os.walk(baseDir + '/accounts'):
|
|
for handle in dirs:
|
|
if not isAccountDir(handle):
|
|
continue
|
|
contactNickname = handle.split('@')[0]
|
|
sharesFilename = baseDir + '/accounts/' + handle + \
|
|
'/' + sharesFileType + '.json'
|
|
if not os.path.isfile(sharesFilename):
|
|
continue
|
|
|
|
sharesJson = loadJson(sharesFilename)
|
|
if not sharesJson:
|
|
continue
|
|
|
|
(resultsExist, currPage, ctr,
|
|
resultStr) = _htmlSharesResult(baseDir, sharesJson, pageNumber,
|
|
resultsPerPage,
|
|
searchStrLowerList,
|
|
currPage, ctr,
|
|
callingDomain, httpPrefix,
|
|
domainFull,
|
|
contactNickname,
|
|
actor, resultsExist,
|
|
searchStrLower, translate,
|
|
sharesFileType)
|
|
sharedItemsForm += resultStr
|
|
|
|
if currPage > pageNumber:
|
|
break
|
|
break
|
|
|
|
# search federated shared items
|
|
if sharesFileType == 'shares':
|
|
catalogsDir = baseDir + '/cache/catalogs'
|
|
else:
|
|
catalogsDir = baseDir + '/cache/wantedItems'
|
|
if currPage <= pageNumber and os.path.isdir(catalogsDir):
|
|
for subdir, dirs, files in os.walk(catalogsDir):
|
|
for f in files:
|
|
if '#' in f:
|
|
continue
|
|
if not f.endswith('.' + sharesFileType + '.json'):
|
|
continue
|
|
federatedDomain = f.split('.')[0]
|
|
if federatedDomain not in sharedItemsFederatedDomains:
|
|
continue
|
|
sharesFilename = catalogsDir + '/' + f
|
|
sharesJson = loadJson(sharesFilename)
|
|
if not sharesJson:
|
|
continue
|
|
|
|
(resultsExist, currPage, ctr,
|
|
resultStr) = _htmlSharesResult(baseDir, sharesJson,
|
|
pageNumber,
|
|
resultsPerPage,
|
|
searchStrLowerList,
|
|
currPage, ctr,
|
|
callingDomain, httpPrefix,
|
|
domainFull,
|
|
contactNickname,
|
|
actor, resultsExist,
|
|
searchStrLower, translate,
|
|
sharesFileType)
|
|
sharedItemsForm += resultStr
|
|
|
|
if currPage > pageNumber:
|
|
break
|
|
break
|
|
|
|
if not resultsExist:
|
|
sharedItemsForm += \
|
|
'<center><h5>' + translate['No results'] + '</h5></center>\n'
|
|
sharedItemsForm += htmlFooter()
|
|
return sharedItemsForm
|
|
|
|
|
|
def htmlSearchEmojiTextEntry(cssCache: {}, translate: {},
|
|
baseDir: str, path: str) -> str:
|
|
"""Search for an emoji by name
|
|
"""
|
|
# emoji.json is generated so that it can be customized and the changes
|
|
# will be retained even if default_emoji.json is subsequently updated
|
|
if not os.path.isfile(baseDir + '/emoji/emoji.json'):
|
|
copyfile(baseDir + '/emoji/default_emoji.json',
|
|
baseDir + '/emoji/emoji.json')
|
|
|
|
actor = path.replace('/search', '')
|
|
domain, port = getDomainFromActor(actor)
|
|
|
|
setCustomBackground(baseDir, 'search-background', 'follow-background')
|
|
|
|
cssFilename = baseDir + '/epicyon-follow.css'
|
|
if os.path.isfile(baseDir + '/follow.css'):
|
|
cssFilename = baseDir + '/follow.css'
|
|
|
|
instanceTitle = \
|
|
getConfigParam(baseDir, 'instanceTitle')
|
|
emojiStr = htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
emojiStr += '<div class="follow">\n'
|
|
emojiStr += ' <div class="followAvatar">\n'
|
|
emojiStr += ' <center>\n'
|
|
emojiStr += \
|
|
' <p class="followText">' + \
|
|
translate['Enter an emoji name to search for'] + '</p>\n'
|
|
emojiStr += ' <form role="search" method="POST" action="' + \
|
|
actor + '/searchhandleemoji">\n'
|
|
emojiStr += ' <input type="hidden" name="actor" value="' + \
|
|
actor + '">\n'
|
|
emojiStr += ' <input type="text" name="searchtext" autofocus><br>\n'
|
|
emojiStr += \
|
|
' <button type="submit" class="button" name="submitSearch">' + \
|
|
translate['Submit'] + '</button>\n'
|
|
emojiStr += ' </form>\n'
|
|
emojiStr += ' </center>\n'
|
|
emojiStr += ' </div>\n'
|
|
emojiStr += '</div>\n'
|
|
emojiStr += htmlFooter()
|
|
return emojiStr
|
|
|
|
|
|
def htmlSearch(cssCache: {}, translate: {},
|
|
baseDir: str, path: str, domain: str,
|
|
defaultTimeline: str, theme: str,
|
|
textModeBanner: str, accessKeys: {}) -> str:
|
|
"""Search called from the timeline icon
|
|
"""
|
|
actor = path.replace('/search', '')
|
|
searchNickname = getNicknameFromActor(actor)
|
|
|
|
setCustomBackground(baseDir, 'search-background', 'follow-background')
|
|
|
|
cssFilename = baseDir + '/epicyon-search.css'
|
|
if os.path.isfile(baseDir + '/search.css'):
|
|
cssFilename = baseDir + '/search.css'
|
|
|
|
instanceTitle = getConfigParam(baseDir, 'instanceTitle')
|
|
followStr = htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
|
|
# show a banner above the search box
|
|
searchBannerFile, searchBannerFilename = \
|
|
getSearchBannerFile(baseDir, searchNickname, domain, theme)
|
|
|
|
textModeBannerStr = htmlKeyboardNavigation(textModeBanner, {}, {})
|
|
if textModeBannerStr is None:
|
|
textModeBannerStr = ''
|
|
|
|
if os.path.isfile(searchBannerFilename):
|
|
timelineKey = accessKeys['menuTimeline']
|
|
usersPath = '/users/' + searchNickname
|
|
followStr += \
|
|
'<header>\n' + textModeBannerStr + \
|
|
'<a href="' + usersPath + '/' + defaultTimeline + '" title="' + \
|
|
translate['Switch to timeline view'] + '" alt="' + \
|
|
translate['Switch to timeline view'] + '" ' + \
|
|
'accesskey="' + timelineKey + '">\n'
|
|
followStr += '<img loading="lazy" class="timeline-banner" src="' + \
|
|
usersPath + '/' + searchBannerFile + '" alt="" /></a>\n' + \
|
|
'</header>\n'
|
|
|
|
# show the search box
|
|
followStr += '<div class="follow">\n'
|
|
followStr += ' <div class="followAvatar">\n'
|
|
followStr += ' <center>\n'
|
|
followStr += \
|
|
' <p class="followText">' + translate['Search screen text'] + '</p>\n'
|
|
followStr += ' <form role="search" method="POST" ' + \
|
|
'accept-charset="UTF-8" action="' + actor + '/searchhandle">\n'
|
|
followStr += \
|
|
' <input type="hidden" name="actor" value="' + actor + '">\n'
|
|
followStr += ' <input type="text" name="searchtext" autofocus><br>\n'
|
|
submitKey = accessKeys['submitButton']
|
|
followStr += ' <button type="submit" class="button" ' + \
|
|
'name="submitSearch" accesskey="' + submitKey + '">' + \
|
|
translate['Submit'] + '</button>\n'
|
|
followStr += ' </form>\n'
|
|
|
|
cachedHashtagSwarmFilename = \
|
|
acctDir(baseDir, searchNickname, domain) + '/.hashtagSwarm'
|
|
swarmStr = ''
|
|
if os.path.isfile(cachedHashtagSwarmFilename):
|
|
try:
|
|
with open(cachedHashtagSwarmFilename, 'r') as fp:
|
|
swarmStr = fp.read()
|
|
except BaseException:
|
|
print('EX: htmlSearch unable to read cached hashtag swarm ' +
|
|
cachedHashtagSwarmFilename)
|
|
pass
|
|
if not swarmStr:
|
|
swarmStr = htmlHashTagSwarm(baseDir, actor, translate)
|
|
if swarmStr:
|
|
try:
|
|
with open(cachedHashtagSwarmFilename, 'w+') as fp:
|
|
fp.write(swarmStr)
|
|
except BaseException:
|
|
print('EX: htmlSearch unable to save cached hashtag swarm ' +
|
|
cachedHashtagSwarmFilename)
|
|
pass
|
|
|
|
followStr += ' <p class="hashtagswarm">' + swarmStr + '</p>\n'
|
|
followStr += ' </center>\n'
|
|
followStr += ' </div>\n'
|
|
followStr += '</div>\n'
|
|
followStr += htmlFooter()
|
|
return followStr
|
|
|
|
|
|
def htmlSkillsSearch(actor: str,
|
|
cssCache: {}, translate: {}, baseDir: str,
|
|
httpPrefix: str,
|
|
skillsearch: str, instanceOnly: bool,
|
|
postsPerPage: int) -> str:
|
|
"""Show a page containing search results for a skill
|
|
"""
|
|
if skillsearch.startswith('*'):
|
|
skillsearch = skillsearch[1:].strip()
|
|
|
|
skillsearch = skillsearch.lower().strip('\n').strip('\r')
|
|
|
|
results = []
|
|
# search instance accounts
|
|
for subdir, dirs, files in os.walk(baseDir + '/accounts/'):
|
|
for f in files:
|
|
if not f.endswith('.json'):
|
|
continue
|
|
if not isAccountDir(f):
|
|
continue
|
|
actorFilename = os.path.join(subdir, f)
|
|
actorJson = loadJson(actorFilename)
|
|
if actorJson:
|
|
if actorJson.get('id') and \
|
|
noOfActorSkills(actorJson) > 0 and \
|
|
actorJson.get('name') and \
|
|
actorJson.get('icon'):
|
|
actor = actorJson['id']
|
|
actorSkillsList = actorJson['hasOccupation']['skills']
|
|
skills = getSkillsFromList(actorSkillsList)
|
|
for skillName, skillLevel in skills.items():
|
|
skillName = skillName.lower()
|
|
if not (skillName in skillsearch or
|
|
skillsearch in skillName):
|
|
continue
|
|
skillLevelStr = str(skillLevel)
|
|
if skillLevel < 100:
|
|
skillLevelStr = '0' + skillLevelStr
|
|
if skillLevel < 10:
|
|
skillLevelStr = '0' + skillLevelStr
|
|
indexStr = \
|
|
skillLevelStr + ';' + actor + ';' + \
|
|
actorJson['name'] + \
|
|
';' + actorJson['icon']['url']
|
|
if indexStr not in results:
|
|
results.append(indexStr)
|
|
break
|
|
if not instanceOnly:
|
|
# search actor cache
|
|
for subdir, dirs, files in os.walk(baseDir + '/cache/actors/'):
|
|
for f in files:
|
|
if not f.endswith('.json'):
|
|
continue
|
|
if not isAccountDir(f):
|
|
continue
|
|
actorFilename = os.path.join(subdir, f)
|
|
cachedActorJson = loadJson(actorFilename)
|
|
if cachedActorJson:
|
|
if cachedActorJson.get('actor'):
|
|
actorJson = cachedActorJson['actor']
|
|
if actorJson.get('id') and \
|
|
noOfActorSkills(actorJson) > 0 and \
|
|
actorJson.get('name') and \
|
|
actorJson.get('icon'):
|
|
actor = actorJson['id']
|
|
actorSkillsList = \
|
|
actorJson['hasOccupation']['skills']
|
|
skills = getSkillsFromList(actorSkillsList)
|
|
for skillName, skillLevel in skills.items():
|
|
skillName = skillName.lower()
|
|
if not (skillName in skillsearch or
|
|
skillsearch in skillName):
|
|
continue
|
|
skillLevelStr = str(skillLevel)
|
|
if skillLevel < 100:
|
|
skillLevelStr = '0' + skillLevelStr
|
|
if skillLevel < 10:
|
|
skillLevelStr = '0' + skillLevelStr
|
|
indexStr = \
|
|
skillLevelStr + ';' + actor + ';' + \
|
|
actorJson['name'] + \
|
|
';' + actorJson['icon']['url']
|
|
if indexStr not in results:
|
|
results.append(indexStr)
|
|
break
|
|
|
|
results.sort(reverse=True)
|
|
|
|
cssFilename = baseDir + '/epicyon-profile.css'
|
|
if os.path.isfile(baseDir + '/epicyon.css'):
|
|
cssFilename = baseDir + '/epicyon.css'
|
|
|
|
instanceTitle = \
|
|
getConfigParam(baseDir, 'instanceTitle')
|
|
skillSearchForm = htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
skillSearchForm += \
|
|
'<center><h1><a href = "' + actor + '/search">' + \
|
|
translate['Skills search'] + ': ' + \
|
|
skillsearch + \
|
|
'</a></h1></center>'
|
|
|
|
if len(results) == 0:
|
|
skillSearchForm += \
|
|
'<center><h5>' + translate['No results'] + \
|
|
'</h5></center>'
|
|
else:
|
|
skillSearchForm += '<center>'
|
|
ctr = 0
|
|
for skillMatch in results:
|
|
skillMatchFields = skillMatch.split(';')
|
|
if len(skillMatchFields) != 4:
|
|
continue
|
|
actor = skillMatchFields[1]
|
|
actorName = skillMatchFields[2]
|
|
avatarUrl = skillMatchFields[3]
|
|
skillSearchForm += \
|
|
'<div class="search-result""><a href="' + \
|
|
actor + '/skills">'
|
|
skillSearchForm += \
|
|
'<img loading="lazy" src="' + avatarUrl + \
|
|
'" alt="" /><span class="search-result-text">' + actorName + \
|
|
'</span></a></div>'
|
|
ctr += 1
|
|
if ctr >= postsPerPage:
|
|
break
|
|
skillSearchForm += '</center>'
|
|
skillSearchForm += htmlFooter()
|
|
return skillSearchForm
|
|
|
|
|
|
def htmlHistorySearch(cssCache: {}, translate: {}, baseDir: str,
|
|
httpPrefix: str,
|
|
nickname: str, domain: str,
|
|
historysearch: str,
|
|
postsPerPage: int, pageNumber: int,
|
|
projectVersion: str,
|
|
recentPostsCache: {},
|
|
maxRecentPosts: int,
|
|
session,
|
|
cachedWebfingers,
|
|
personCache: {},
|
|
port: int,
|
|
YTReplacementDomain: str,
|
|
twitterReplacementDomain: str,
|
|
showPublishedDateOnly: bool,
|
|
peertubeInstances: [],
|
|
allowLocalNetworkAccess: bool,
|
|
themeName: str, boxName: str,
|
|
systemLanguage: str,
|
|
maxLikeCount: int,
|
|
signingPrivateKeyPem: str,
|
|
CWlists: {},
|
|
listsEnabled: str) -> str:
|
|
"""Show a page containing search results for your post history
|
|
"""
|
|
if historysearch.startswith("'"):
|
|
historysearch = historysearch[1:].strip()
|
|
|
|
historysearch = historysearch.lower().strip('\n').strip('\r')
|
|
|
|
boxFilenames = \
|
|
searchBoxPosts(baseDir, nickname, domain,
|
|
historysearch, postsPerPage, boxName)
|
|
|
|
cssFilename = baseDir + '/epicyon-profile.css'
|
|
if os.path.isfile(baseDir + '/epicyon.css'):
|
|
cssFilename = baseDir + '/epicyon.css'
|
|
|
|
instanceTitle = \
|
|
getConfigParam(baseDir, 'instanceTitle')
|
|
historySearchForm = \
|
|
htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
|
|
# add the page title
|
|
domainFull = getFullDomain(domain, port)
|
|
actor = localActorUrl(httpPrefix, nickname, domainFull)
|
|
historySearchTitle = '🔍 ' + translate['Your Posts']
|
|
if boxName == 'bookmarks':
|
|
historySearchTitle = '🔍 ' + translate['Bookmarks']
|
|
|
|
historySearchForm += \
|
|
'<center><h1><a href="' + actor + '/search">' + \
|
|
historySearchTitle + '</a></h1></center>'
|
|
|
|
if len(boxFilenames) == 0:
|
|
historySearchForm += \
|
|
'<center><h5>' + translate['No results'] + \
|
|
'</h5></center>'
|
|
return historySearchForm
|
|
|
|
separatorStr = htmlPostSeparator(baseDir, None)
|
|
|
|
# ensure that the page number is in bounds
|
|
if not pageNumber:
|
|
pageNumber = 1
|
|
elif pageNumber < 1:
|
|
pageNumber = 1
|
|
|
|
# get the start end end within the index file
|
|
startIndex = int((pageNumber - 1) * postsPerPage)
|
|
endIndex = startIndex + postsPerPage
|
|
noOfBoxFilenames = len(boxFilenames)
|
|
if endIndex >= noOfBoxFilenames and noOfBoxFilenames > 0:
|
|
endIndex = noOfBoxFilenames - 1
|
|
|
|
index = startIndex
|
|
while index <= endIndex:
|
|
postFilename = boxFilenames[index]
|
|
if not postFilename:
|
|
index += 1
|
|
continue
|
|
postJsonObject = loadJson(postFilename)
|
|
if not postJsonObject:
|
|
index += 1
|
|
continue
|
|
showIndividualPostIcons = True
|
|
allowDeletion = False
|
|
postStr = \
|
|
individualPostAsHtml(signingPrivateKeyPem,
|
|
True, recentPostsCache,
|
|
maxRecentPosts,
|
|
translate, None,
|
|
baseDir, session, cachedWebfingers,
|
|
personCache,
|
|
nickname, domain, port,
|
|
postJsonObject,
|
|
None, True, allowDeletion,
|
|
httpPrefix, projectVersion,
|
|
'search',
|
|
YTReplacementDomain,
|
|
twitterReplacementDomain,
|
|
showPublishedDateOnly,
|
|
peertubeInstances,
|
|
allowLocalNetworkAccess,
|
|
themeName, systemLanguage, maxLikeCount,
|
|
showIndividualPostIcons,
|
|
showIndividualPostIcons,
|
|
False, False, False, False,
|
|
CWlists, listsEnabled)
|
|
if postStr:
|
|
historySearchForm += separatorStr + postStr
|
|
index += 1
|
|
|
|
historySearchForm += htmlFooter()
|
|
return historySearchForm
|
|
|
|
|
|
def htmlHashtagSearch(cssCache: {},
|
|
nickname: str, domain: str, port: int,
|
|
recentPostsCache: {}, maxRecentPosts: int,
|
|
translate: {},
|
|
baseDir: str, hashtag: str, pageNumber: int,
|
|
postsPerPage: int,
|
|
session, cachedWebfingers: {}, personCache: {},
|
|
httpPrefix: str, projectVersion: str,
|
|
YTReplacementDomain: str,
|
|
twitterReplacementDomain: str,
|
|
showPublishedDateOnly: bool,
|
|
peertubeInstances: [],
|
|
allowLocalNetworkAccess: bool,
|
|
themeName: str, systemLanguage: str,
|
|
maxLikeCount: int,
|
|
signingPrivateKeyPem: str,
|
|
CWlists: {}, listsEnabled: str) -> str:
|
|
"""Show a page containing search results for a hashtag
|
|
or after selecting a hashtag from the swarm
|
|
"""
|
|
if hashtag.startswith('#'):
|
|
hashtag = hashtag[1:]
|
|
hashtag = urllib.parse.unquote(hashtag)
|
|
hashtagIndexFile = baseDir + '/tags/' + hashtag + '.txt'
|
|
if not os.path.isfile(hashtagIndexFile):
|
|
if hashtag != hashtag.lower():
|
|
hashtag = hashtag.lower()
|
|
hashtagIndexFile = baseDir + '/tags/' + hashtag + '.txt'
|
|
if not os.path.isfile(hashtagIndexFile):
|
|
print('WARN: hashtag file not found ' + hashtagIndexFile)
|
|
return None
|
|
|
|
separatorStr = htmlPostSeparator(baseDir, None)
|
|
|
|
# check that the directory for the nickname exists
|
|
if nickname:
|
|
accountDir = acctDir(baseDir, nickname, domain)
|
|
if not os.path.isdir(accountDir):
|
|
nickname = None
|
|
|
|
# read the index
|
|
with open(hashtagIndexFile, 'r') as f:
|
|
lines = f.readlines()
|
|
|
|
# read the css
|
|
cssFilename = baseDir + '/epicyon-profile.css'
|
|
if os.path.isfile(baseDir + '/epicyon.css'):
|
|
cssFilename = baseDir + '/epicyon.css'
|
|
|
|
# ensure that the page number is in bounds
|
|
if not pageNumber:
|
|
pageNumber = 1
|
|
elif pageNumber < 1:
|
|
pageNumber = 1
|
|
|
|
# get the start end end within the index file
|
|
startIndex = int((pageNumber - 1) * postsPerPage)
|
|
endIndex = startIndex + postsPerPage
|
|
noOfLines = len(lines)
|
|
if endIndex >= noOfLines and noOfLines > 0:
|
|
endIndex = noOfLines - 1
|
|
|
|
# add the page title
|
|
instanceTitle = \
|
|
getConfigParam(baseDir, 'instanceTitle')
|
|
hashtagSearchForm = \
|
|
htmlHeaderWithExternalStyle(cssFilename, instanceTitle)
|
|
if nickname:
|
|
hashtagSearchForm += '<center>\n' + \
|
|
'<h1><a href="/users/' + nickname + '/search">#' + \
|
|
hashtag + '</a></h1>\n'
|
|
else:
|
|
hashtagSearchForm += '<center>\n' + \
|
|
'<h1>#' + hashtag + '</h1>\n'
|
|
|
|
# RSS link for hashtag feed
|
|
hashtagSearchForm += '<a href="/tags/rss2/' + hashtag + '">'
|
|
hashtagSearchForm += \
|
|
'<img style="width:3%;min-width:50px" ' + \
|
|
'loading="lazy" alt="RSS 2.0" title="RSS 2.0" src="/' + \
|
|
'icons/logorss.png" /></a></center>\n'
|
|
|
|
# edit the category for this hashtag
|
|
if isEditor(baseDir, nickname):
|
|
category = getHashtagCategory(baseDir, hashtag)
|
|
hashtagSearchForm += '<div class="hashtagCategoryContainer">\n'
|
|
hashtagSearchForm += ' <form enctype="multipart/form-data" ' + \
|
|
'method="POST" accept-charset="UTF-8" action="' + \
|
|
'/users/' + nickname + '/tags/' + hashtag + \
|
|
'/sethashtagcategory">\n'
|
|
hashtagSearchForm += ' <center>\n'
|
|
hashtagSearchForm += translate['Category']
|
|
hashtagSearchForm += \
|
|
' <input type="text" style="width: 20ch" ' + \
|
|
'name="hashtagCategory" value="' + category + '">\n'
|
|
hashtagSearchForm += \
|
|
' <button type="submit" class="button" name="submitYes">' + \
|
|
translate['Submit'] + '</button>\n'
|
|
hashtagSearchForm += ' </center>\n'
|
|
hashtagSearchForm += ' </form>\n'
|
|
hashtagSearchForm += '</div>\n'
|
|
|
|
if startIndex > 0:
|
|
# previous page link
|
|
hashtagSearchForm += \
|
|
' <center>\n' + \
|
|
' <a href="/users/' + nickname + \
|
|
'/tags/' + hashtag + '?page=' + \
|
|
str(pageNumber - 1) + \
|
|
'"><img loading="lazy" class="pageicon" src="/' + \
|
|
'icons/pageup.png" title="' + \
|
|
translate['Page up'] + \
|
|
'" alt="' + translate['Page up'] + \
|
|
'"></a>\n </center>\n'
|
|
index = startIndex
|
|
while index <= endIndex:
|
|
postId = lines[index].strip('\n').strip('\r')
|
|
if ' ' not in postId:
|
|
nickname = getNicknameFromActor(postId)
|
|
if not nickname:
|
|
index += 1
|
|
continue
|
|
else:
|
|
postFields = postId.split(' ')
|
|
if len(postFields) != 3:
|
|
index += 1
|
|
continue
|
|
nickname = postFields[1]
|
|
postId = postFields[2]
|
|
postFilename = locatePost(baseDir, nickname, domain, postId)
|
|
if not postFilename:
|
|
index += 1
|
|
continue
|
|
postJsonObject = loadJson(postFilename)
|
|
if not postJsonObject:
|
|
index += 1
|
|
continue
|
|
if not isPublicPost(postJsonObject):
|
|
index += 1
|
|
continue
|
|
showIndividualPostIcons = False
|
|
if nickname:
|
|
showIndividualPostIcons = True
|
|
allowDeletion = False
|
|
showRepeats = showIndividualPostIcons
|
|
showIcons = showIndividualPostIcons
|
|
manuallyApprovesFollowers = False
|
|
showPublicOnly = False
|
|
storeToCache = False
|
|
allowDownloads = True
|
|
avatarUrl = None
|
|
showAvatarOptions = True
|
|
postStr = \
|
|
individualPostAsHtml(signingPrivateKeyPem,
|
|
allowDownloads, recentPostsCache,
|
|
maxRecentPosts,
|
|
translate, None,
|
|
baseDir, session, cachedWebfingers,
|
|
personCache,
|
|
nickname, domain, port,
|
|
postJsonObject,
|
|
avatarUrl, showAvatarOptions,
|
|
allowDeletion,
|
|
httpPrefix, projectVersion,
|
|
'search',
|
|
YTReplacementDomain,
|
|
twitterReplacementDomain,
|
|
showPublishedDateOnly,
|
|
peertubeInstances,
|
|
allowLocalNetworkAccess,
|
|
themeName, systemLanguage, maxLikeCount,
|
|
showRepeats, showIcons,
|
|
manuallyApprovesFollowers,
|
|
showPublicOnly,
|
|
storeToCache, False, CWlists,
|
|
listsEnabled)
|
|
if postStr:
|
|
hashtagSearchForm += separatorStr + postStr
|
|
index += 1
|
|
|
|
if endIndex < noOfLines - 1:
|
|
# next page link
|
|
hashtagSearchForm += \
|
|
' <center>\n' + \
|
|
' <a href="/users/' + nickname + '/tags/' + hashtag + \
|
|
'?page=' + str(pageNumber + 1) + \
|
|
'"><img loading="lazy" class="pageicon" src="/icons' + \
|
|
'/pagedown.png" title="' + translate['Page down'] + \
|
|
'" alt="' + translate['Page down'] + '"></a>' + \
|
|
' </center>'
|
|
hashtagSearchForm += htmlFooter()
|
|
return hashtagSearchForm
|
|
|
|
|
|
def rssHashtagSearch(nickname: str, domain: str, port: int,
|
|
recentPostsCache: {}, maxRecentPosts: int,
|
|
translate: {},
|
|
baseDir: str, hashtag: str,
|
|
postsPerPage: int,
|
|
session, cachedWebfingers: {}, personCache: {},
|
|
httpPrefix: str, projectVersion: str,
|
|
YTReplacementDomain: str,
|
|
twitterReplacementDomain: str,
|
|
systemLanguage: str) -> str:
|
|
"""Show an rss feed for a hashtag
|
|
"""
|
|
if hashtag.startswith('#'):
|
|
hashtag = hashtag[1:]
|
|
hashtag = urllib.parse.unquote(hashtag)
|
|
hashtagIndexFile = baseDir + '/tags/' + hashtag + '.txt'
|
|
if not os.path.isfile(hashtagIndexFile):
|
|
if hashtag != hashtag.lower():
|
|
hashtag = hashtag.lower()
|
|
hashtagIndexFile = baseDir + '/tags/' + hashtag + '.txt'
|
|
if not os.path.isfile(hashtagIndexFile):
|
|
print('WARN: hashtag file not found ' + hashtagIndexFile)
|
|
return None
|
|
|
|
# check that the directory for the nickname exists
|
|
if nickname:
|
|
accountDir = acctDir(baseDir, nickname, domain)
|
|
if not os.path.isdir(accountDir):
|
|
nickname = None
|
|
|
|
# read the index
|
|
lines = []
|
|
with open(hashtagIndexFile, 'r') as f:
|
|
lines = f.readlines()
|
|
if not lines:
|
|
return None
|
|
|
|
domainFull = getFullDomain(domain, port)
|
|
|
|
maxFeedLength = 10
|
|
hashtagFeed = \
|
|
rss2TagHeader(hashtag, httpPrefix, domainFull)
|
|
for index in range(len(lines)):
|
|
postId = lines[index].strip('\n').strip('\r')
|
|
if ' ' not in postId:
|
|
nickname = getNicknameFromActor(postId)
|
|
if not nickname:
|
|
index += 1
|
|
if index >= maxFeedLength:
|
|
break
|
|
continue
|
|
else:
|
|
postFields = postId.split(' ')
|
|
if len(postFields) != 3:
|
|
index += 1
|
|
if index >= maxFeedLength:
|
|
break
|
|
continue
|
|
nickname = postFields[1]
|
|
postId = postFields[2]
|
|
postFilename = locatePost(baseDir, nickname, domain, postId)
|
|
if not postFilename:
|
|
index += 1
|
|
if index >= maxFeedLength:
|
|
break
|
|
continue
|
|
postJsonObject = loadJson(postFilename)
|
|
if postJsonObject:
|
|
if not isPublicPost(postJsonObject):
|
|
index += 1
|
|
if index >= maxFeedLength:
|
|
break
|
|
continue
|
|
# add to feed
|
|
if postJsonObject['object'].get('content') and \
|
|
postJsonObject['object'].get('attributedTo') and \
|
|
postJsonObject['object'].get('published'):
|
|
published = postJsonObject['object']['published']
|
|
pubDate = datetime.strptime(published, "%Y-%m-%dT%H:%M:%SZ")
|
|
rssDateStr = pubDate.strftime("%a, %d %b %Y %H:%M:%S UT")
|
|
hashtagFeed += ' <item>'
|
|
hashtagFeed += \
|
|
' <author>' + \
|
|
postJsonObject['object']['attributedTo'] + \
|
|
'</author>'
|
|
if postJsonObject['object'].get('summary'):
|
|
hashtagFeed += \
|
|
' <title>' + \
|
|
postJsonObject['object']['summary'] + \
|
|
'</title>'
|
|
description = \
|
|
getBaseContentFromPost(postJsonObject, systemLanguage)
|
|
description = firstParagraphFromString(description)
|
|
hashtagFeed += \
|
|
' <description>' + description + '</description>'
|
|
hashtagFeed += \
|
|
' <pubDate>' + rssDateStr + '</pubDate>'
|
|
if postJsonObject['object'].get('attachment'):
|
|
for attach in postJsonObject['object']['attachment']:
|
|
if not attach.get('url'):
|
|
continue
|
|
hashtagFeed += \
|
|
' <link>' + attach['url'] + '</link>'
|
|
hashtagFeed += ' </item>'
|
|
index += 1
|
|
if index >= maxFeedLength:
|
|
break
|
|
|
|
return hashtagFeed + rss2TagFooter()
|