Use schema metadata only if it matches features

[lhoestq]

e.g. if we use map in arrow format and transform the table, the returned table might have new columns but the metadata might be wrong

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[github-actions]

Show benchmarks

PyArrow==8.0.0

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

metric	read_batch_formatted_as_numpy after write_array2d	read_batch_formatted_as_numpy after write_flattened_sequence	read_batch_formatted_as_numpy after write_nested_sequence	read_batch_unformated after write_array2d	read_batch_unformated after write_flattened_sequence	read_batch_unformated after write_nested_sequence	read_col_formatted_as_numpy after write_array2d	read_col_formatted_as_numpy after write_flattened_sequence	read_col_formatted_as_numpy after write_nested_sequence	read_col_unformated after write_array2d	read_col_unformated after write_flattened_sequence	read_col_unformated after write_nested_sequence	read_formatted_as_numpy after write_array2d	read_formatted_as_numpy after write_flattened_sequence	read_formatted_as_numpy after write_nested_sequence	read_unformated after write_array2d	read_unformated after write_flattened_sequence	read_unformated after write_nested_sequence	write_array2d	write_flattened_sequence	write_nested_sequence
new / old (diff)	0.005382 / 0.011353 (-0.005970)	0.003853 / 0.011008 (-0.007155)	0.062629 / 0.038508 (0.024121)	0.030344 / 0.023109 (0.007234)	0.245394 / 0.275898 (-0.030505)	0.266004 / 0.323480 (-0.057476)	0.003183 / 0.007986 (-0.004802)	0.002795 / 0.004328 (-0.001533)	0.048357 / 0.004250 (0.044107)	0.043834 / 0.037052 (0.006782)	0.255979 / 0.258489 (-0.002510)	0.280803 / 0.293841 (-0.013038)	0.028200 / 0.128546 (-0.100347)	0.010856 / 0.075646 (-0.064791)	0.207076 / 0.419271 (-0.212195)	0.036286 / 0.043533 (-0.007247)	0.246492 / 0.255139 (-0.008647)	0.265861 / 0.283200 (-0.017338)	0.018309 / 0.141683 (-0.123374)	1.155136 / 1.452155 (-0.297018)	1.214342 / 1.492716 (-0.278375)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.092530 / 0.018006 (0.074524)	0.344951 / 0.000490 (0.344461)	0.000207 / 0.000200 (0.000007)	0.000043 / 0.000054 (-0.000011)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.018324 / 0.037411 (-0.019087)	0.063137 / 0.014526 (0.048611)	0.074683 / 0.176557 (-0.101874)	0.120224 / 0.737135 (-0.616912)	0.083107 / 0.296338 (-0.213232)

Benchmark: benchmark_iterating.json

metric	read 5000	read 50000	read_batch 50000 10	read_batch 50000 100	read_batch 50000 1000	read_formatted numpy 5000	read_formatted pandas 5000	read_formatted tensorflow 5000	read_formatted torch 5000	read_formatted_batch numpy 5000 10	read_formatted_batch numpy 5000 1000	shuffled read 5000	shuffled read 50000	shuffled read_batch 50000 10	shuffled read_batch 50000 100	shuffled read_batch 50000 1000	shuffled read_formatted numpy 5000	shuffled read_formatted_batch numpy 5000 10	shuffled read_formatted_batch numpy 5000 1000
new / old (diff)	0.288631 / 0.215209 (0.073422)	2.817992 / 2.077655 (0.740337)	1.473609 / 1.504120 (-0.030511)	1.336610 / 1.541195 (-0.204585)	1.354807 / 1.468490 (-0.113683)	0.568776 / 4.584777 (-4.016001)	2.412607 / 3.745712 (-1.333105)	2.832816 / 5.269862 (-2.437045)	1.789899 / 4.565676 (-2.775778)	0.063602 / 0.424275 (-0.360673)	0.004993 / 0.007607 (-0.002615)	0.338830 / 0.226044 (0.112786)	3.302550 / 2.268929 (1.033621)	1.827907 / 55.444624 (-53.616717)	1.589857 / 6.876477 (-5.286620)	1.647746 / 2.142072 (-0.494326)	0.658461 / 4.805227 (-4.146766)	0.120360 / 6.500664 (-6.380304)	0.042989 / 0.075469 (-0.032480)

Benchmark: benchmark_map_filter.json

metric	filter	map fast-tokenizer batched	map identity	map identity batched	map no-op batched	map no-op batched numpy	map no-op batched pandas	map no-op batched pytorch	map no-op batched tensorflow
new / old (diff)	0.945487 / 1.841788 (-0.896301)	11.846335 / 8.074308 (3.772027)	10.483199 / 10.191392 (0.291807)	0.131853 / 0.680424 (-0.548570)	0.014230 / 0.534201 (-0.519971)	0.288700 / 0.579283 (-0.290584)	0.276086 / 0.434364 (-0.158278)	0.326225 / 0.540337 (-0.214112)	0.422874 / 1.386936 (-0.964062)

PyArrow==latest

Show updated benchmarks!

Benchmark: benchmark_array_xd.json

metric	read_batch_formatted_as_numpy after write_array2d	read_batch_formatted_as_numpy after write_flattened_sequence	read_batch_formatted_as_numpy after write_nested_sequence	read_batch_unformated after write_array2d	read_batch_unformated after write_flattened_sequence	read_batch_unformated after write_nested_sequence	read_col_formatted_as_numpy after write_array2d	read_col_formatted_as_numpy after write_flattened_sequence	read_col_formatted_as_numpy after write_nested_sequence	read_col_unformated after write_array2d	read_col_unformated after write_flattened_sequence	read_col_unformated after write_nested_sequence	read_formatted_as_numpy after write_array2d	read_formatted_as_numpy after write_flattened_sequence	read_formatted_as_numpy after write_nested_sequence	read_unformated after write_array2d	read_unformated after write_flattened_sequence	read_unformated after write_nested_sequence	write_array2d	write_flattened_sequence	write_nested_sequence
new / old (diff)	0.006234 / 0.011353 (-0.005118)	0.004104 / 0.011008 (-0.006904)	0.049967 / 0.038508 (0.011459)	0.037157 / 0.023109 (0.014048)	0.261892 / 0.275898 (-0.014006)	0.284304 / 0.323480 (-0.039176)	0.004482 / 0.007986 (-0.003504)	0.002920 / 0.004328 (-0.001409)	0.048827 / 0.004250 (0.044577)	0.052258 / 0.037052 (0.015206)	0.277121 / 0.258489 (0.018632)	0.304177 / 0.293841 (0.010336)	0.053537 / 0.128546 (-0.075009)	0.011137 / 0.075646 (-0.064509)	0.058188 / 0.419271 (-0.361083)	0.034283 / 0.043533 (-0.009250)	0.261912 / 0.255139 (0.006773)	0.273851 / 0.283200 (-0.009348)	0.017824 / 0.141683 (-0.123859)	1.130454 / 1.452155 (-0.321701)	1.176834 / 1.492716 (-0.315882)

Benchmark: benchmark_getitem_100B.json

metric	get_batch_of_1024_random_rows	get_batch_of_1024_rows	get_first_row	get_last_row
new / old (diff)	0.102104 / 0.018006 (0.084098)	0.302873 / 0.000490 (0.302383)	0.000208 / 0.000200 (0.000008)	0.000052 / 0.000054 (-0.000003)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.022470 / 0.037411 (-0.014941)	0.076776 / 0.014526 (0.062250)	0.088220 / 0.176557 (-0.088337)	0.130030 / 0.737135 (-0.607105)	0.089955 / 0.296338 (-0.206383)

Benchmark: benchmark_iterating.json

metric	read 5000	read 50000	read_batch 50000 10	read_batch 50000 100	read_batch 50000 1000	read_formatted numpy 5000	read_formatted pandas 5000	read_formatted tensorflow 5000	read_formatted torch 5000	read_formatted_batch numpy 5000 10	read_formatted_batch numpy 5000 1000	shuffled read 5000	shuffled read 50000	shuffled read_batch 50000 10	shuffled read_batch 50000 100	shuffled read_batch 50000 1000	shuffled read_formatted numpy 5000	shuffled read_formatted_batch numpy 5000 10	shuffled read_formatted_batch numpy 5000 1000
new / old (diff)	0.284070 / 0.215209 (0.068861)	2.769130 / 2.077655 (0.691475)	1.546379 / 1.504120 (0.042259)	1.435849 / 1.541195 (-0.105346)	1.478616 / 1.468490 (0.010126)	0.569185 / 4.584777 (-4.015592)	2.504721 / 3.745712 (-1.240992)	2.778267 / 5.269862 (-2.491595)	1.860360 / 4.565676 (-2.705316)	0.073465 / 0.424275 (-0.350810)	0.005108 / 0.007607 (-0.002499)	0.335185 / 0.226044 (0.109140)	3.314799 / 2.268929 (1.045870)	1.934824 / 55.444624 (-53.509801)	1.656247 / 6.876477 (-5.220229)	1.785422 / 2.142072 (-0.356650)	0.673677 / 4.805227 (-4.131551)	0.117692 / 6.500664 (-6.382972)	0.041648 / 0.075469 (-0.033821)

Benchmark: benchmark_map_filter.json

metric	filter	map fast-tokenizer batched	map identity	map identity batched	map no-op batched	map no-op batched numpy	map no-op batched pandas	map no-op batched pytorch	map no-op batched tensorflow
new / old (diff)	0.972143 / 1.841788 (-0.869645)	12.980353 / 8.074308 (4.906045)	11.056189 / 10.191392 (0.864797)	0.134592 / 0.680424 (-0.545832)	0.015972 / 0.534201 (-0.518229)	0.301691 / 0.579283 (-0.277593)	0.286332 / 0.434364 (-0.148032)	0.329025 / 0.540337 (-0.211312)	0.422585 / 1.386936 (-0.964351)