288 lines
10 KiB
Python
288 lines
10 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.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, Tuple, 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", "400")) # safety cap
|
|
|
|
def dprint(msg: str) -> None:
|
|
if DIAG:
|
|
print(f"[SCRAPER] {msg}")
|
|
|
|
def _resolve_target(req: ScrapeRequest) -> int:
|
|
# accept both "count" and "max_count"
|
|
return int((getattr(req, "count", None) if getattr(req, "count", None) is not None else req.max_count))
|
|
|
|
def _db_existing_ids(db: Session, ig_ids: List[str]) -> Set[str]:
|
|
if not ig_ids:
|
|
return set()
|
|
rows = (
|
|
db.query(InstagramPost.ig_post_id)
|
|
.filter(InstagramPost.ig_post_id.in_(ig_ids))
|
|
.all()
|
|
)
|
|
return {r[0] for r in rows}
|
|
|
|
def _load_posts_by_ids_ordered(db: Session, account_id, ig_ids: List[str]) -> List[InstagramPost]:
|
|
"""Fetch already-saved posts for this account, preserving ig_ids order."""
|
|
if not ig_ids:
|
|
return []
|
|
rows = (
|
|
db.query(InstagramPost)
|
|
.filter(InstagramPost.account_id == account_id, InstagramPost.ig_post_id.in_(ig_ids))
|
|
.all()
|
|
)
|
|
by_id = {r.ig_post_id: r for r in rows}
|
|
return [by_id[i] for i in ig_ids if i in by_id]
|
|
|
|
def _normalize_page(page: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
|
|
if not page:
|
|
return []
|
|
return parse_boxapi_items_fast(page)
|
|
|
|
def _collect_normalized_pages(
|
|
db: Session,
|
|
username: str,
|
|
target: int,
|
|
start_cursor: Optional[str],
|
|
since_ts: Optional[int],
|
|
) -> Tuple[List[Dict[str, Any]], List[str], Optional[str], int]:
|
|
"""
|
|
Collect pages while counting ONLY *new* posts toward `target`.
|
|
Keep dup IDs to optionally top-up from DB after paging.
|
|
Returns:
|
|
normalized_new -> normalized items not in DB (in order found)
|
|
dup_ids_ordered -> ig_post_id list already in DB (in order found)
|
|
final_cursor -> last next_max_id seen (persist to continue older)
|
|
pages_walked -> how many API pages we pulled
|
|
"""
|
|
normalized_new: List[Dict[str, Any]] = []
|
|
dup_ids_ordered: List[str] = []
|
|
pages = 0
|
|
|
|
# Start from stored cursor if present to keep drilling older across runs
|
|
max_id = start_cursor
|
|
newest_cursor = None
|
|
final_cursor: Optional[str] = None
|
|
seen_cursors: Set[Optional[str]] = set()
|
|
|
|
while len(normalized_new) < target and pages < MAX_PAGES:
|
|
per_req = 12 # BoxAPI cap per request
|
|
try:
|
|
page, _log, nxt = get_media_page(username=username, count=per_req, max_id=max_id)
|
|
except BoxAPIError as e:
|
|
dprint(f"BoxAPI error: {e}")
|
|
break
|
|
|
|
pages += 1
|
|
dprint(f"[PAGE]#{pages}: request max_id={max_id!r} → got {len(page)} items, next_max_id={nxt!r}")
|
|
if newest_cursor is None:
|
|
newest_cursor = nxt
|
|
|
|
if not page:
|
|
dprint("No items → stopping.")
|
|
break
|
|
|
|
norm = _normalize_page(page)
|
|
|
|
# optional time filter
|
|
if since_ts is not None:
|
|
try:
|
|
s = int(since_ts)
|
|
norm = [n for n in norm if (n.get("taken_at") or 0) > s]
|
|
except Exception:
|
|
pass
|
|
|
|
ids = [str(n.get("ig_post_id")) for n in norm if n.get("ig_post_id")]
|
|
already = _db_existing_ids(db, ids)
|
|
new_here = [n for n in norm if n.get("ig_post_id") and str(n["ig_post_id"]) not in already]
|
|
dups_here = [str(n["ig_post_id"]) for n in norm if n.get("ig_post_id") and str(n["ig_post_id"]) in already]
|
|
|
|
# Append in discovery order
|
|
normalized_new.extend(new_here)
|
|
dup_ids_ordered.extend(dups_here)
|
|
|
|
dprint(
|
|
f" normalized={len(norm)} new={len(new_here)} dups={len(dups_here)} "
|
|
f"new_collected_total={len(normalized_new)}/{target}"
|
|
)
|
|
|
|
# Stop only when we have enough NEW items
|
|
if len(normalized_new) >= target:
|
|
final_cursor = nxt # where to resume next time (older side)
|
|
break
|
|
|
|
# Advance to older page; if none, we're done
|
|
if not nxt:
|
|
dprint("No next_max_id → end of feed")
|
|
final_cursor = None
|
|
break
|
|
|
|
# Loop guard: avoid cursor cycles
|
|
if nxt in seen_cursors:
|
|
dprint("Repeating cursor detected → stopping.")
|
|
final_cursor = nxt
|
|
break
|
|
|
|
seen_cursors.add(nxt)
|
|
final_cursor = nxt
|
|
max_id = nxt # CRITICAL: go older
|
|
|
|
return normalized_new, dup_ids_ordered, final_cursor, pages
|
|
|
|
|
|
def scrape_and_store(db: Session, req: ScrapeRequest) -> List[ScrapedPost]:
|
|
target = _resolve_target(req)
|
|
dprint(f"requested count={getattr(req,'count',None)} max_count={req.max_count} → resolved target={target}")
|
|
|
|
# since_taken_at (optional)
|
|
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
|
|
|
|
# Ensure account row (needed for DB top-up)
|
|
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)
|
|
|
|
# Start from stored cursor (for deeper backfill)
|
|
start_cursor = get_cursor(db, req.username)
|
|
dprint(f"start_cursor={start_cursor!r}")
|
|
|
|
# 1) Collect normalized candidates across pages
|
|
normalized_new, dup_ids_ordered, final_cursor, pages_walked = _collect_normalized_pages(
|
|
db=db,
|
|
username=req.username,
|
|
target=target,
|
|
start_cursor=start_cursor,
|
|
since_ts=since_ts,
|
|
)
|
|
|
|
# Persist backfill cursor only if we actually traversed older pages
|
|
if pages_walked > 0 and final_cursor is not None:
|
|
set_cursor(db, req.username, final_cursor)
|
|
|
|
result: List[ScrapedPost] = []
|
|
|
|
# 2) First, save NEW posts (upload + insert) until we reach target or run out
|
|
for item in normalized_new:
|
|
if len(result) >= target:
|
|
break
|
|
|
|
post_id = str(item["ig_post_id"])
|
|
# Skip if somehow inserted by a parallel worker
|
|
if db.query(InstagramPost).filter_by(ig_post_id=post_id).first():
|
|
continue
|
|
|
|
seller_id = str(req.seller_id)
|
|
media_urls = list(item.get("remote_urls", []))
|
|
|
|
# If video, keep only video URLs (avoid thumb)
|
|
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
|
|
|
|
created_at = datetime.fromtimestamp(item["taken_at"]) if item.get("taken_at") else datetime.utcnow()
|
|
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}")
|
|
|
|
row = 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(row)
|
|
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,
|
|
))
|
|
|
|
# 3) If we still need more, TOP-UP with already-saved posts from DB (NO uploads)
|
|
if len(result) < target and dup_ids_ordered:
|
|
missing = target - len(result)
|
|
# Load the dup rows in the same order as we saw them while paging
|
|
dup_rows = _load_posts_by_ids_ordered(db, account.id, dup_ids_ordered)
|
|
for row in dup_rows:
|
|
if missing <= 0:
|
|
break
|
|
# Build response directly from DB (fast)
|
|
result.append(ScrapedPost(
|
|
ig_post_id=row.ig_post_id,
|
|
media_type=row.media_type,
|
|
caption=row.caption,
|
|
thumbnail_url=row.thumbnail_url,
|
|
local_media=row.local_media or [],
|
|
created_at=row.created_at,
|
|
))
|
|
missing -= 1
|
|
|
|
# 4) If STILL not enough (end of feed + tiny DB), just return what we have
|
|
if len(result) < target:
|
|
dprint(f"feed_exhausted_or_insufficient_total=True returned={len(result)} requested={target}")
|
|
|
|
return result
|