forked from Mirrors/itch-dl
Add filtering job URLs by regex/globs
This commit is contained in:
@@ -54,6 +54,12 @@ def parse_args() -> argparse.Namespace:
|
||||
help="filter downloaded files with a shell-style glob/fnmatch (unmatched files are skipped)")
|
||||
parser.add_argument("--filter-files-regex", metavar="regex", default=None,
|
||||
help="filter downloaded files with a Python regex (unmatched files are skipped)")
|
||||
|
||||
parser.add_argument("--filter-urls-glob", metavar="glob", default=None,
|
||||
help="filter itch URLs with a shell-style glob/fnmatch (unmatched URLs are skipped)")
|
||||
parser.add_argument("--filter-urls-regex", metavar="regex", default=None,
|
||||
help="filter itch URLs with a Python regex (unmatched URLs are skipped)")
|
||||
|
||||
parser.add_argument("--verbose", action="store_true",
|
||||
help="print verbose logs")
|
||||
|
||||
@@ -89,8 +95,10 @@ def run() -> int:
|
||||
)
|
||||
|
||||
jobs = get_jobs_for_url_or_path(url_or_path, settings)
|
||||
jobs = list(set(jobs)) # Deduplicate, just in case...
|
||||
logging.info("Found %d URL(s).", len(jobs))
|
||||
logging.info("Found %d URL(s) total.", len(jobs))
|
||||
|
||||
jobs = preprocess_job_urls(jobs, settings)
|
||||
logging.info("Will process %d URL(s) after filtering and deduplication.", len(jobs))
|
||||
|
||||
if len(jobs) == 0:
|
||||
sys.exit("No URLs to download.")
|
||||
|
||||
Reference in New Issue
Block a user