parameterize dump images step to work with custom data paths

crawler/3_dump_images.py

@@ -1,11 +1,12 @@
 import json
+import pathlib
 from urllib.request import urlretrieve
 from tqdm import tqdm
 from data_access import Listing
 
 
-def dump_images():
-    for listing in tqdm(Listing.get_all_listings()):
+def dump_images(listing_paths: list[str]):
+    for listing in tqdm(Listing.get_all_listings(listing_paths)):
         with open(listing.path_detail_json()) as f:
             detail = json.load(f)
 
@@ -29,12 +30,13 @@ def dump_images():
             tqdm.write(str(p))
             try:
                 urlretrieve(url, p)
-            except:
-                tqdm.write(f"404 for {url}")
+            except Exception as e:
+                tqdm.write(f"Error for {url}: {e}")
 
 
 def main():
-    dump_images()
+    listing_paths = sorted(list(pathlib.Path("data/rs").glob("*/listing.json")))
+    dump_images(listing_paths)
 
 
 if __name__ == "__main__":

crawler/data_access.py

@@ -19,7 +19,12 @@ class Listing:
         for listing_path in listing_paths:
             with open(listing_path) as f:
                 d = json.load(f)
-            identifiers.append(Listing(d["identifier"]))
+
+            # data_dir is the first directory before the listing_path
+            data_dir = pathlib.Path(listing_path)
+            while str(d['identifier']) in str(data_dir.resolve().absolute()):
+                data_dir = data_dir.parent
+            identifiers.append(Listing(d["identifier"], data_dir=data_dir))
 
         return identifiers
 

crawler/main.py

@@ -66,9 +66,12 @@ def dump_detail(ctx: click.core.Context):
 
 
 @cli.command()
-def dump_images():
-    click.echo('Running dump_images')
-    dump_images_module.dump_images()
+@click.pass_context
+def dump_images(ctx: click.core.Context):
+    data_dir = ctx.obj['data_dir']
+    click.echo(f'Running dump_images stored in {data_dir}')
+    listing_paths = sorted(list(pathlib.Path(data_dir).glob("*/listing.json")))
+    dump_images_module.dump_images(listing_paths)
 
 
 @cli.command()