Enforce identical Viewer feature types

#9
Files changed (3) hide show
  1. CHANGELOG.md +3 -0
  2. README.md +118 -0
  3. VALIDATION.md +5 -0
CHANGELOG.md CHANGED
@@ -19,6 +19,9 @@
19
  - Replaced the combined `default/train` surface with separate
20
  `canonical/archive` and `pre_standard/archive` configurations.
21
  - Made `canonical` the default configuration.
 
 
 
22
  - Declared stored dimensions and byte sizes as integers and `has_audio` as a
23
  boolean; retained floating-point frame rate for historical encodings.
24
  - Continued to exclude the nested provenance subtree through direct-file
 
19
  - Replaced the combined `default/train` surface with separate
20
  `canonical/archive` and `pre_standard/archive` configurations.
21
  - Made `canonical` the default configuration.
22
+ - Declared the full feature contract inside each builder configuration so the
23
+ generated Parquet schemas remain identical even when a source CSV column
24
+ happens to contain only integer-looking values.
25
  - Declared stored dimensions and byte sizes as integers and `has_audio` as a
26
  boolean; retained floating-point frame rate for historical encodings.
27
  - Continued to exclude the nested provenance subtree through direct-file
README.md CHANGED
@@ -151,12 +151,130 @@ dataset_info:
151
  configs:
152
  - config_name: canonical
153
  default: true
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
154
  data_files:
155
  - split: archive
156
  path:
157
  - metadata.csv
158
  - Standard_Time_Lapses/*.mp4
159
  - config_name: pre_standard
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
160
  data_files:
161
  - split: archive
162
  path:
 
151
  configs:
152
  - config_name: canonical
153
  default: true
154
+ features:
155
+ - name: video
156
+ dtype: video
157
+ - name: record_id
158
+ dtype: large_string
159
+ - name: collection_tier
160
+ dtype: large_string
161
+ - name: is_canonical
162
+ dtype: bool
163
+ - name: acquisition_protocol_version
164
+ dtype: large_string
165
+ - name: date
166
+ dtype: large_string
167
+ - name: time
168
+ dtype: large_string
169
+ - name: tool
170
+ dtype: large_string
171
+ - name: medium
172
+ dtype: large_string
173
+ - name: support
174
+ dtype: large_string
175
+ - name: dimensions
176
+ dtype: large_string
177
+ - name: duration_seconds
178
+ dtype: float64
179
+ - name: width_px
180
+ dtype: int64
181
+ - name: height_px
182
+ dtype: int64
183
+ - name: orientation
184
+ dtype: large_string
185
+ - name: encoded_frame_rate
186
+ dtype: float64
187
+ - name: video_codec
188
+ dtype: large_string
189
+ - name: pixel_format
190
+ dtype: large_string
191
+ - name: container
192
+ dtype: large_string
193
+ - name: has_audio
194
+ dtype: bool
195
+ - name: hub_xet_hash
196
+ dtype: large_string
197
+ - name: original_filename
198
+ dtype: large_string
199
+ - name: provenance_status
200
+ dtype: large_string
201
+ - name: legacy_provenance_path
202
+ dtype: large_string
203
+ - name: legacy_related_references
204
+ dtype: large_string
205
+ - name: evidence_method
206
+ dtype: large_string
207
+ - name: provenance_reference_duration_delta_seconds
208
+ dtype: float64
209
+ - name: size_bytes
210
+ dtype: int64
211
+ - name: notes
212
+ dtype: large_string
213
  data_files:
214
  - split: archive
215
  path:
216
  - metadata.csv
217
  - Standard_Time_Lapses/*.mp4
218
  - config_name: pre_standard
219
+ features:
220
+ - name: video
221
+ dtype: video
222
+ - name: record_id
223
+ dtype: large_string
224
+ - name: collection_tier
225
+ dtype: large_string
226
+ - name: is_canonical
227
+ dtype: bool
228
+ - name: acquisition_protocol_version
229
+ dtype: large_string
230
+ - name: date
231
+ dtype: large_string
232
+ - name: time
233
+ dtype: large_string
234
+ - name: tool
235
+ dtype: large_string
236
+ - name: medium
237
+ dtype: large_string
238
+ - name: support
239
+ dtype: large_string
240
+ - name: dimensions
241
+ dtype: large_string
242
+ - name: duration_seconds
243
+ dtype: float64
244
+ - name: width_px
245
+ dtype: int64
246
+ - name: height_px
247
+ dtype: int64
248
+ - name: orientation
249
+ dtype: large_string
250
+ - name: encoded_frame_rate
251
+ dtype: float64
252
+ - name: video_codec
253
+ dtype: large_string
254
+ - name: pixel_format
255
+ dtype: large_string
256
+ - name: container
257
+ dtype: large_string
258
+ - name: has_audio
259
+ dtype: bool
260
+ - name: hub_xet_hash
261
+ dtype: large_string
262
+ - name: original_filename
263
+ dtype: large_string
264
+ - name: provenance_status
265
+ dtype: large_string
266
+ - name: legacy_provenance_path
267
+ dtype: large_string
268
+ - name: legacy_related_references
269
+ dtype: large_string
270
+ - name: evidence_method
271
+ dtype: large_string
272
+ - name: provenance_reference_duration_delta_seconds
273
+ dtype: float64
274
+ - name: size_bytes
275
+ dtype: int64
276
+ - name: notes
277
+ dtype: large_string
278
  data_files:
279
  - split: archive
280
  path:
VALIDATION.md CHANGED
@@ -35,6 +35,11 @@ content identities were read from the Hub repository inventory.
35
  | Stored-orientation conflicts | 0 |
36
  | Duplicate active content hashes | 0 |
37
 
 
 
 
 
 
38
  ## Audited statistics
39
 
40
  - Total indexed playback duration: 50.06 hours
 
35
  | Stored-orientation conflicts | 0 |
36
  | Duplicate active content hashes | 0 |
37
 
38
+ Both Hub builder configurations explicitly declare the same 29 features. This
39
+ prevents per-CSV inference from representing an integer-only canonical
40
+ `encoded_frame_rate` column as `int64` while representing historical
41
+ fractional rates as `float64` in the Pre-Standard configuration.
42
+
43
  ## Audited statistics
44
 
45
  - Total indexed playback duration: 50.06 hours