mirror of
https://github.com/ArchiveBox/ArchiveBox
synced 2024-11-22 20:23:12 +00:00
68 lines
2.6 KiB
Python
68 lines
2.6 KiB
Python
__package__ = 'archivebox.seeds'
|
|
|
|
import abx
|
|
|
|
from django.utils.html import format_html_join, format_html
|
|
|
|
from abid_utils.admin import ABIDModelAdmin
|
|
|
|
from archivebox import DATA_DIR
|
|
|
|
from seeds.models import Seed
|
|
|
|
|
|
|
|
class SeedAdmin(ABIDModelAdmin):
|
|
list_display = ('abid', 'created_at', 'created_by', 'label', 'notes', 'uri', 'extractor', 'tags_str', 'crawls', 'num_crawls', 'num_snapshots')
|
|
sort_fields = ('abid', 'created_at', 'created_by', 'label', 'notes', 'uri', 'extractor', 'tags_str')
|
|
search_fields = ('abid', 'created_by__username', 'label', 'notes', 'uri', 'extractor', 'tags_str')
|
|
|
|
readonly_fields = ('created_at', 'modified_at', 'abid_info', 'scheduled_crawls', 'crawls', 'snapshots', 'contents')
|
|
fields = ('label', 'notes', 'uri', 'extractor', 'tags_str', 'config', 'created_by', *readonly_fields)
|
|
|
|
list_filter = ('extractor', 'created_by')
|
|
ordering = ['-created_at']
|
|
list_per_page = 100
|
|
actions = ["delete_selected"]
|
|
|
|
def num_crawls(self, obj):
|
|
return obj.crawl_set.count()
|
|
|
|
def num_snapshots(self, obj):
|
|
return obj.snapshot_set.count()
|
|
|
|
def scheduled_crawls(self, obj):
|
|
return format_html_join('<br/>', ' - <a href="{}">{}</a>', (
|
|
(scheduledcrawl.admin_change_url, scheduledcrawl)
|
|
for scheduledcrawl in obj.scheduled_crawl_set.all().order_by('-created_at')[:20]
|
|
)) or format_html('<i>No Scheduled Crawls yet...</i>')
|
|
|
|
def crawls(self, obj):
|
|
return format_html_join('<br/>', ' - <a href="{}">{}</a>', (
|
|
(crawl.admin_change_url, crawl)
|
|
for crawl in obj.crawl_set.all().order_by('-created_at')[:20]
|
|
)) or format_html('<i>No Crawls yet...</i>')
|
|
|
|
def snapshots(self, obj):
|
|
return format_html_join('<br/>', ' - <a href="{}">{}</a>', (
|
|
(snapshot.admin_change_url, snapshot)
|
|
for snapshot in obj.snapshot_set.all().order_by('-created_at')[:20]
|
|
)) or format_html('<i>No Snapshots yet...</i>')
|
|
|
|
def contents(self, obj):
|
|
if obj.uri.startswith('file:///data/'):
|
|
source_file = DATA_DIR / obj.uri.replace('file:///data/', '', 1)
|
|
contents = ""
|
|
try:
|
|
contents = source_file.read_text().strip()[:14_000]
|
|
except Exception as e:
|
|
contents = f'Error reading {source_file}: {e}'
|
|
|
|
return format_html('<b><code>{}</code>:</b><br/><pre>{}</pre>', source_file, contents)
|
|
|
|
return format_html('See URLs here: <a href="{}">{}</a>', obj.uri, obj.uri)
|
|
|
|
|
|
@abx.hookimpl
|
|
def register_admin(admin_site):
|
|
admin_site.register(Seed, SeedAdmin)
|