274 lines
9.5 KiB
Python
274 lines
9.5 KiB
Python
# scraper.py
|
|
|
|
import os
|
|
from sqlalchemy.orm import Session
|
|
from database import InstagramAccount, InstagramPost
|
|
from utils.boxapi_client import get_media_page, BoxAPIError
|
|
from utils.parser_fast import parse_boxapi_items_fast
|
|
from utils.seen_store import filter_unseen, mark_seen
|
|
from utils.s3_uploader import (
|
|
upload_media_from_url,
|
|
upload_image_from_url_optimized,
|
|
upload_video_from_url_optimized,
|
|
)
|
|
from schemas import ScrapeRequest, ScrapedPost
|
|
from typing import List, Dict, Any, Optional, Set
|
|
from datetime import datetime
|
|
from scraper_cursor import get_cursor, set_cursor
|
|
|
|
DIAG = os.getenv("SCRAPER_DIAGNOSTICS", "0") == "1"
|
|
MAX_PAGES = int(os.getenv("SCRAPER_MAX_PAGES", "200")) # hard safety
|
|
|
|
def dprint(msg: str) -> None:
|
|
if DIAG:
|
|
print(f"[SCRAPER] {msg}")
|
|
|
|
def _ig_id_of(it: Dict[str, Any]) -> Optional[str]:
|
|
v = it.get("id") or it.get("pk")
|
|
return str(v) if v is not None else None
|
|
|
|
def _taken_at_of(it: Dict[str, Any]) -> Optional[int]:
|
|
v = it.get("taken_at") or it.get("timestamp")
|
|
try:
|
|
return int(v) if v is not None else None
|
|
except Exception:
|
|
return None
|
|
|
|
def _db_existing_ids(db: Session, ids: List[str]) -> Set[str]:
|
|
if not ids:
|
|
return set()
|
|
rows = (
|
|
db.query(InstagramPost.ig_post_id)
|
|
.filter(InstagramPost.ig_post_id.in_(ids))
|
|
.all()
|
|
)
|
|
return {r[0] for r in rows}
|
|
|
|
def _ordered_unseen_ids(db: Session, username: str, ids_in_order: List[str]) -> List[str]:
|
|
"""
|
|
Combine Redis fast-path + DB guard to get truly unseen ids in original order.
|
|
This eliminates mismatch issues between Redis key format and DB.
|
|
"""
|
|
# Redis fast path
|
|
redis_unseen = set(filter_unseen(username, ids_in_order))
|
|
if not redis_unseen:
|
|
return []
|
|
|
|
# DB guard (in case posts were stored by another worker/process)
|
|
already_in_db = _db_existing_ids(db, ids_in_order)
|
|
truly_unseen = redis_unseen.difference(already_in_db)
|
|
|
|
# Preserve original order
|
|
return [pid for pid in ids_in_order if pid in truly_unseen]
|
|
|
|
def _collect_unseen_raw(
|
|
db: Session,
|
|
username: str,
|
|
target: int,
|
|
stored_cursor: Optional[str],
|
|
since_ts: Optional[int],
|
|
) -> tuple[List[Dict[str, Any]], Optional[str], int, int]:
|
|
"""
|
|
Page through BoxAPI until we collect `target` unseen posts or run out of pages.
|
|
Returns: (raw_unseen_items, final_cursor, pages_walked, unseen_count)
|
|
"""
|
|
raw_unseen: List[Dict[str, Any]] = []
|
|
pages_walked = 0
|
|
final_cursor = stored_cursor # default to stored; will update when we traverse
|
|
first = True
|
|
cursor = stored_cursor # For first fetch we will use None (newest) unless you want pure backfill.
|
|
|
|
while len(raw_unseen) < target and pages_walked < MAX_PAGES:
|
|
max_id = None if first else cursor # newest first, then backfill using cursor
|
|
need = min(12, max(1, target - len(raw_unseen)))
|
|
|
|
try:
|
|
page, _log, next_id = get_media_page(username=username, count=need, max_id=max_id)
|
|
except BoxAPIError as e:
|
|
dprint(f"BoxAPI error on page fetch: {e}")
|
|
break
|
|
|
|
pages_walked += 1
|
|
dprint(f"Page#{pages_walked} fetch → max_id={max_id} → next_max_id={next_id}, raw_count={len(page)}")
|
|
|
|
# Move traversal cursor forward only if we actually fetched a page
|
|
cursor = next_id
|
|
if page:
|
|
final_cursor = next_id
|
|
|
|
if not page:
|
|
break
|
|
|
|
# Apply since_ts BEFORE unseen checks
|
|
filtered = []
|
|
ids_in_order: List[str] = []
|
|
for it in page:
|
|
igid = _ig_id_of(it)
|
|
if not igid:
|
|
continue
|
|
ts = _taken_at_of(it)
|
|
if since_ts is not None and ts is not None and ts <= since_ts:
|
|
continue
|
|
filtered.append(it)
|
|
ids_in_order.append(igid)
|
|
|
|
if not ids_in_order:
|
|
dprint(" after since_ts filter: 0 items")
|
|
if not cursor:
|
|
break
|
|
first = False
|
|
continue
|
|
|
|
# Combined unseen check (Redis + DB)
|
|
ordered_unseen_ids = _ordered_unseen_ids(db, username, ids_in_order)
|
|
dprint(f" ids_in_order={len(ids_in_order)} unseen_after_redis+db={len(ordered_unseen_ids)}")
|
|
|
|
if ordered_unseen_ids:
|
|
unseen_set = set(ordered_unseen_ids)
|
|
for it in filtered:
|
|
igid = _ig_id_of(it)
|
|
if igid in unseen_set:
|
|
raw_unseen.append(it)
|
|
if len(raw_unseen) >= target:
|
|
break
|
|
|
|
if not cursor:
|
|
break
|
|
|
|
first = False
|
|
|
|
dprint(f"Collected unseen={len(raw_unseen)} across pages={pages_walked}, final_cursor={final_cursor}")
|
|
return raw_unseen, final_cursor, pages_walked, len(raw_unseen)
|
|
|
|
def scrape_and_store(db: Session, req: ScrapeRequest) -> List[ScrapedPost]:
|
|
target = int(req.max_count)
|
|
|
|
# Parse optional since_taken_at
|
|
since_ts: Optional[int] = None
|
|
if hasattr(req, "since_taken_at") and req.since_taken_at is not None:
|
|
try:
|
|
since_ts = int(req.since_taken_at)
|
|
except Exception:
|
|
since_ts = None
|
|
|
|
stored_cursor = get_cursor(db, req.username) # may be None
|
|
|
|
# 1) Collect ONLY unseen raw items (newest page first, then older via cursor)
|
|
raw_unseen, final_cursor, pages_walked, unseen_found = _collect_unseen_raw(
|
|
db=db,
|
|
username=req.username,
|
|
target=target,
|
|
stored_cursor=stored_cursor,
|
|
since_ts=since_ts,
|
|
)
|
|
|
|
# Persist traversal progress only if we actually walked at least one page
|
|
# (This avoids accidentally overwriting a good cursor when nothing fetched.)
|
|
if pages_walked > 0:
|
|
set_cursor(db, req.username, final_cursor)
|
|
|
|
if not raw_unseen:
|
|
dprint("No unseen items collected; returning early.")
|
|
return []
|
|
|
|
# 2) Normalize only selected unseen raw items
|
|
normalized = parse_boxapi_items_fast(raw_unseen)
|
|
|
|
# 3) Defensive unseen re-check (race safety) AFTER normalize
|
|
ids_in_order = [n.get("ig_post_id") for n in normalized if n.get("ig_post_id")]
|
|
final_unseen_ids = _ordered_unseen_ids(db, req.username, ids_in_order)
|
|
batch = [n for n in normalized if n.get("ig_post_id") in set(final_unseen_ids)][:target]
|
|
|
|
# Mark seen to prevent reprocessing
|
|
mark_seen(req.username, [b["ig_post_id"] for b in batch])
|
|
|
|
# 4) Ensure account row
|
|
account = db.query(InstagramAccount).filter_by(username=req.username).first()
|
|
if not account:
|
|
account = InstagramAccount(
|
|
username=req.username,
|
|
seller_id=req.seller_id,
|
|
is_active=True,
|
|
last_synced=None
|
|
)
|
|
db.add(account)
|
|
db.commit()
|
|
db.refresh(account)
|
|
|
|
# 5) Upload + save each post
|
|
result: List[ScrapedPost] = []
|
|
for item in batch:
|
|
post_id = str(item["ig_post_id"])
|
|
seller_id = str(req.seller_id)
|
|
|
|
# DB guard
|
|
if db.query(InstagramPost).filter_by(ig_post_id=post_id).first():
|
|
continue
|
|
|
|
media_urls = list(item.get("remote_urls", []))
|
|
|
|
if item["media_type"] == "video":
|
|
thumb = (item.get("thumbnail_url") or "").strip()
|
|
def is_video(u: str) -> bool:
|
|
base = u.split("?", 1)[0].lower()
|
|
return base.endswith((".mp4", ".mov", ".m4v", ".webm"))
|
|
media_urls = [u for u in media_urls if is_video(u) and u != thumb]
|
|
|
|
uploaded_urls: List[str] = []
|
|
for remote_url in media_urls:
|
|
try:
|
|
if item["media_type"] == "video":
|
|
url = upload_video_from_url_optimized(remote_url, seller_id, post_id)
|
|
elif item["media_type"] in ("image", "carousel"):
|
|
url = upload_image_from_url_optimized(remote_url, seller_id, post_id)
|
|
if url is None:
|
|
url = upload_media_from_url(remote_url, seller_id, post_id)
|
|
else:
|
|
url = upload_media_from_url(remote_url, seller_id, post_id)
|
|
if url:
|
|
uploaded_urls.append(url)
|
|
except Exception as e:
|
|
print(f"⚠️ Error uploading media: {e}")
|
|
continue
|
|
|
|
final_thumbnail = item.get("thumbnail_url") or None
|
|
if final_thumbnail:
|
|
try:
|
|
thumb_url = upload_image_from_url_optimized(final_thumbnail, seller_id, post_id)
|
|
if thumb_url:
|
|
final_thumbnail = thumb_url
|
|
else:
|
|
alt = upload_media_from_url(final_thumbnail, seller_id, post_id)
|
|
if alt:
|
|
final_thumbnail = alt
|
|
except Exception as e:
|
|
print(f"⚠️ Failed to upload thumbnail: {e}")
|
|
|
|
created_at = datetime.fromtimestamp(item["taken_at"]) if item.get("taken_at") else datetime.utcnow()
|
|
post_entry = InstagramPost(
|
|
ig_post_id=post_id,
|
|
account_id=account.id,
|
|
media_type=item["media_type"],
|
|
caption=item.get("caption"),
|
|
thumbnail_url=final_thumbnail,
|
|
local_media=uploaded_urls,
|
|
created_at=created_at,
|
|
)
|
|
db.add(post_entry)
|
|
db.commit()
|
|
|
|
result.append(ScrapedPost(
|
|
ig_post_id=post_id,
|
|
media_type=item["media_type"],
|
|
caption=item.get("caption"),
|
|
thumbnail_url=final_thumbnail,
|
|
local_media=uploaded_urls,
|
|
created_at=created_at,
|
|
))
|
|
|
|
if len(result) >= target:
|
|
break
|
|
|
|
dprint(f"Returning {len(result)} items (requested {target}).")
|
|
return result
|