collective/mediawiki_dump_tools
18
0
Fork 0
Code Issues Pull Requests Activity

support .jsonl.d

Browse Source
This commit is contained in:
Nathan TeBlunthuis
2025-12-22 20:13:04 -08:00
parent 618c343898
commit d822085698
2 changed files with 20 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
src/wikiq/__init__.py
View File

@@ -515,6 +515,7 @@ class WikiqParser:
namespaces: Union[list[int], None] = None,
revert_radius: int = 15,
output_jsonl: bool = False,
output_jsonl_dir: bool = False,
output_parquet: bool = False,
batch_size: int = 1024,
resume_point: Union[tuple, dict, None] = None,
@@ -569,6 +570,7 @@ class WikiqParser:
# Initialize output
self.batch_size = batch_size
self.output_jsonl = output_jsonl
self.output_jsonl_dir = output_jsonl_dir
self.output_parquet = output_parquet
self.output_file = output_file
@@ -836,7 +838,19 @@ class WikiqParser:
part_numbers[ns] = 0
elif self.output_jsonl:
append_mode = self.resume_point is not None
writer = JSONLWriter(self.output_file, schema, append=append_mode)
if self.output_jsonl_dir:
# Create directory for JSONL output
Path(self.output_file).mkdir(parents=True, exist_ok=True)
part_num = 0
if self.resume_point is not None and len(self.resume_point) > 2:
part_num = self.resume_point[2]
part_numbers[None] = part_num
jsonl_path = self._get_part_path(
Path(self.output_file) / "data.jsonl", part_num
)
writer = JSONLWriter(str(jsonl_path), schema, append=append_mode)
else:
writer = JSONLWriter(self.output_file, schema, append=append_mode)
else:
writer = pacsv.CSVWriter(
self.output_file,
@@ -1437,7 +1451,8 @@ def main():
output = "."
# Detect output format from extension
output_jsonl = output.endswith(".jsonl") or output.endswith(".jsonl.d")
output_jsonl_dir = output.endswith(".jsonl.d")
output_jsonl = output.endswith(".jsonl") or output_jsonl_dir
output_parquet = output.endswith(".parquet")
partition_namespaces = args.partition_namespaces and output_parquet
@@ -1495,6 +1510,7 @@ def main():
text=args.text,
diff=args.diff,
output_jsonl=output_jsonl,
output_jsonl_dir=output_jsonl_dir,
output_parquet=output_parquet,
partition_namespaces=partition_namespaces,
batch_size=args.batch_size,