Prepare 0.25.0 release

Author: kindly

Cargo.lock

@@ -1,6 +1,6 @@
 [package]
 name = "flatterer"
-version = "0.24.1"
+version = "0.25.0"
 authors = ["David Raznick <kindly@gmail.com>"]
 edition = "2024"
 license = "MIT"
@@ -14,7 +14,7 @@ serde_json = { version = "1.0.150", features = ["preserve_order"] }
 pyo3 = { version = "0.28.3", features = ["extension-module", "eyre"] }
 eyre = "0.6.12"
 #libflatterer={path = "../libflatterer"}
-libflatterer = "0.24.1"
+libflatterer = "0.25.0"
 
 flatterer-web = "0.24.1"
 #flatterer-web={path = "../flatterer-web"}

Cargo.toml

@@ -4,6 +4,23 @@ All notable changes to this project will be documented in this file.
 
 and this project adheres to [Semantic Versioning](http://semver.org/).
 
+## [0.25.0] - 2026-06-28
+
+### Changed
+- Upgrade `libflatterer` to 0.25.0.
+- Python API and CLI now expose `schema_only`/`--schema-only` and `stream_wrapped`/`--stream-wrapped`.
+- JSON flattening now uses the new jiter streaming parser from `libflatterer`, replacing the previous YAJL parser path.
+- Expect 1.5-2x speed improvements for most JSON shapes in both single-threaded and multithreaded modes.
+- For nested lists of dictionaries, multithreaded runs can be 3-4x faster because the jiter parser is much faster.
+- Wrapped JSON objects now stream arrays of objects by default, reducing memory use for large wrapped inputs.
+- JSON stream input now uses the jiter streaming parser, improving streaming behavior while preserving NDJSON-compatible output.
+- Schema-only output now uses the new structural scan path where supported, avoiding full row materialization.
+- Output cell handling now keeps short strings in `SmartString`, reducing allocation overhead in the hot flattening path.
+- One-to-many link level handling has been simplified in `libflatterer`.
+
+### Fixed
+- Mixed arrays whose first item is an object are now handled consistently, including non-object elements as `value` rows instead of regressing in multicore processing.
+
 ## [0.24.1] - 2026-06-02
 
 ### Fixed
@@ -430,4 +447,3 @@ and this project adheres to [Semantic Versioning](http://semver.org/).
 ### Added
 
 - [Inline One to One option](https://flatterer.opendata.coop/options.html#inline-one-to-one) to mean that if an array only has one item in for all the data then treat it as sub-object.
-

docs/changelog.md

@@ -67,6 +67,9 @@ Options:
   --stats                     Produce stats about the data in the
                               datapackage.json file
   --all-strings               Convert all fields to strings
+  --schema-only               Only output schema files, not data files
+  --stream-wrapped / --no-stream-wrapped
+                              Stream wrapped JSON objects, default true
   --help                      Show this message and exit.
 ```
 
@@ -888,4 +891,40 @@ import flatterer
 
 flatterer.flatten('inputfile.json', 'ouput_dir', all_strings=True)
 ```
+
+## Schema Only
+
+Only output schema files (`datapackage.json`, `fields.csv`, and `tables.csv`) and skip data files.
+
+### CLI Usage
+
+```bash
+flatterer INPUT_FILE OUTPUT_DIRECTORY --schema-only
+```
+
+### Python Usage
+
+```python
+import flatterer
+
+flatterer.flatten('inputfile.json', 'ouput_dir', schema_only=True)
+```
+
+## Stream Wrapped
+
+Stream arrays of objects inside wrapped JSON objects without buffering the whole document. This is enabled by default.
+
+### CLI Usage
+
+```bash
+flatterer INPUT_FILE OUTPUT_DIRECTORY --no-stream-wrapped
+```
+
+### Python Usage
+
+```python
+import flatterer
+
+flatterer.flatten('inputfile.json', 'ouput_dir', stream_wrapped=False)
+```
 

docs/options.md

@@ -99,7 +99,9 @@ def flatten(
     json_path="",
     arrays_new_table=False,
     truncate=False,
-    all_strings=False
+    all_strings=False,
+    schema_only=False,
+    stream_wrapped=True
 ):
     global LOGGING_SETUP
     if not LOGGING_SETUP:
@@ -148,7 +150,7 @@ def flatten(
                        schema, schema_titles, path, json_stream, ndjson, 
                        sqlite_path, threads, log_error, postgres, postgres_schema, 
                        drop, pushdown, sql_scripts, evolve, no_link, stats, low_disk, low_memory, 
-                       gzip_input, json_path, arrays_new_table, truncate, all_strings)
+                       gzip_input, json_path, arrays_new_table, truncate, all_strings, schema_only, stream_wrapped)
         elif method == 'iter':
             if path:
                 raise AttributeError("path not allowed when supplying an iterator")
@@ -161,7 +163,8 @@ def flatten(
                        table_prefix, id_prefix, emit_obj, force,  
                        schema, schema_titles, sqlite_path, threads, log_error, 
                        postgres, postgres_schema, drop, pushdown, sql_scripts, evolve, 
-                       no_link, stats, low_disk, low_memory, gzip_input, json_path, arrays_new_table, truncate, all_strings)
+                       no_link, stats, low_disk, low_memory, gzip_input, json_path, arrays_new_table, truncate, all_strings,
+                       schema_only, stream_wrapped)
         else:
             raise AttributeError("input needs to be a string or a generator of strings, dicts or bytes")
 
@@ -175,7 +178,7 @@ def flatten(
             data=PrettyDict()
         )
 
-        if csv:
+        if csv and not schema_only:
             for name, title in output['tables'].values:
                 csv_path = os.path.join(output_dir, 'csv', str(title) + '.csv')
                 if dataframe:
@@ -187,13 +190,13 @@ def flatten(
                 else:
                     output['data'][title] = csv_path
 
-        if sqlite:
+        if sqlite and not schema_only:
             output['sqlite'] = os.path.join(output_dir, 'sqlite.db')
 
-        if xlsx:
+        if xlsx and not schema_only:
             output['xlsx'] = os.path.join(output_dir, 'output.xlsx')
 
-        if ods:
+        if ods and not schema_only:
             output['ods'] = os.path.join(output_dir, 'output.ods')
 
         return output
@@ -257,6 +260,8 @@ def iterator_flatten(*args, **kw):
 @click.option('--id-prefix', default="", help='Prefix for all `_link` id fields')
 @click.option('--stats', is_flag=True, default=False, help='Produce stats about the data in the datapackage.json file')
 @click.option('--all-strings', is_flag=True, default=False, help='Convert all fields to strings')
+@click.option('--schema-only', is_flag=True, default=False, help='Only output schema files, not data files')
+@click.option('--stream-wrapped/--no-stream-wrapped', default=True, help='Stream wrapped JSON objects, default true')
 @click.argument('inputs', required=False, nargs=-1)
 @click.argument('output_directory', required=False)
 def cli(
@@ -296,7 +301,9 @@ def cli(
     json_path="",
     arrays_new_table=False,
     truncate=False,
-    all_strings=False
+    all_strings=False,
+    schema_only=False,
+    stream_wrapped=True
 ):
     if web:
         import pathlib
@@ -370,6 +377,8 @@ def cli(
                 json_path=json_path,
                 arrays_new_table=arrays_new_table,
                 truncate=truncate,
-                all_strings=all_strings)
+                all_strings=all_strings,
+                schema_only=schema_only,
+                stream_wrapped=stream_wrapped)
     except IOError:
         pass