Fix regression for pandas < 2.0.0 in JSON loader

[albertvillanova]

A regression was introduced for pandas < 2.0.0 in PR:

• Preserve JSON column order and support list of strings field #6914

As described in pandas docs, the dtype_backend parameter was first added in pandas 2.0.0: https://pandas.pydata.org/docs/reference/api/pandas.read_json.html

This PR fixes the regression by passing (or not) the dtype_backend parameter depending on pandas version.

Maybe, in a future 3.0 datasets release, we could just require pandas > 2.0.

Reported by:

• load json file error with v2.20.0 #6977

[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)	0.005144 / 0.011353 (-0.006209)	0.003500 / 0.011008 (-0.007509)	0.063670 / 0.038508 (0.025162)	0.031793 / 0.023109 (0.008683)	0.239611 / 0.275898 (-0.036287)	0.276681 / 0.323480 (-0.046799)	0.004148 / 0.007986 (-0.003838)	0.002713 / 0.004328 (-0.001615)	0.048832 / 0.004250 (0.044582)	0.043066 / 0.037052 (0.006014)	0.256835 / 0.258489 (-0.001655)	0.292224 / 0.293841 (-0.001617)	0.027530 / 0.128546 (-0.101017)	0.010509 / 0.075646 (-0.065137)	0.203370 / 0.419271 (-0.215901)	0.035643 / 0.043533 (-0.007890)	0.252161 / 0.255139 (-0.002978)	0.271883 / 0.283200 (-0.011316)	0.018658 / 0.141683 (-0.123024)	1.081676 / 1.452155 (-0.370479)	1.142146 / 1.492716 (-0.350571)

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)	0.093484 / 0.018006 (0.075477)	0.298607 / 0.000490 (0.298117)	0.000220 / 0.000200 (0.000020)	0.000044 / 0.000054 (-0.000010)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.019021 / 0.037411 (-0.018390)	0.062471 / 0.014526 (0.047946)	0.075393 / 0.176557 (-0.101163)	0.121040 / 0.737135 (-0.616095)	0.077613 / 0.296338 (-0.218726)

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)	0.294857 / 0.215209 (0.079648)	2.931143 / 2.077655 (0.853489)	1.510866 / 1.504120 (0.006746)	1.379574 / 1.541195 (-0.161621)	1.352358 / 1.468490 (-0.116133)	0.561670 / 4.584777 (-4.023107)	2.378434 / 3.745712 (-1.367278)	2.713203 / 5.269862 (-2.556658)	1.706416 / 4.565676 (-2.859260)	0.062355 / 0.424275 (-0.361920)	0.004971 / 0.007607 (-0.002636)	0.336498 / 0.226044 (0.110453)	3.316464 / 2.268929 (1.047535)	1.833035 / 55.444624 (-53.611589)	1.532808 / 6.876477 (-5.343668)	1.537323 / 2.142072 (-0.604749)	0.639430 / 4.805227 (-4.165798)	0.115808 / 6.500664 (-6.384856)	0.043545 / 0.075469 (-0.031924)

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)	0.974428 / 1.841788 (-0.867360)	11.368914 / 8.074308 (3.294606)	9.754488 / 10.191392 (-0.436904)	0.146277 / 0.680424 (-0.534146)	0.013917 / 0.534201 (-0.520284)	0.286809 / 0.579283 (-0.292474)	0.267144 / 0.434364 (-0.167219)	0.326161 / 0.540337 (-0.214177)	0.418059 / 1.386936 (-0.968877)

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)	0.005341 / 0.011353 (-0.006012)	0.003460 / 0.011008 (-0.007548)	0.050135 / 0.038508 (0.011627)	0.032014 / 0.023109 (0.008905)	0.259835 / 0.275898 (-0.016063)	0.286275 / 0.323480 (-0.037205)	0.004350 / 0.007986 (-0.003636)	0.002800 / 0.004328 (-0.001529)	0.049358 / 0.004250 (0.045107)	0.040182 / 0.037052 (0.003130)	0.278352 / 0.258489 (0.019863)	0.307869 / 0.293841 (0.014028)	0.029151 / 0.128546 (-0.099395)	0.010091 / 0.075646 (-0.065555)	0.058814 / 0.419271 (-0.360458)	0.033150 / 0.043533 (-0.010383)	0.263594 / 0.255139 (0.008455)	0.284065 / 0.283200 (0.000866)	0.017968 / 0.141683 (-0.123714)	1.145605 / 1.452155 (-0.306550)	1.196884 / 1.492716 (-0.295832)

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)	0.094045 / 0.018006 (0.076039)	0.299031 / 0.000490 (0.298541)	0.000210 / 0.000200 (0.000011)	0.000044 / 0.000054 (-0.000010)

Benchmark: benchmark_indices_mapping.json

metric	select	shard	shuffle	sort	train_test_split
new / old (diff)	0.022510 / 0.037411 (-0.014901)	0.077478 / 0.014526 (0.062953)	0.087746 / 0.176557 (-0.088811)	0.129311 / 0.737135 (-0.607825)	0.089921 / 0.296338 (-0.206418)

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)	0.290279 / 0.215209 (0.075070)	2.880725 / 2.077655 (0.803070)	1.541262 / 1.504120 (0.037142)	1.424475 / 1.541195 (-0.116719)	1.436397 / 1.468490 (-0.032093)	0.578237 / 4.584777 (-4.006540)	0.965249 / 3.745712 (-2.780463)	2.682534 / 5.269862 (-2.587327)	1.732859 / 4.565676 (-2.832817)	0.065523 / 0.424275 (-0.358752)	0.005466 / 0.007607 (-0.002141)	0.343985 / 0.226044 (0.117940)	3.397463 / 2.268929 (1.128534)	1.929370 / 55.444624 (-53.515255)	1.605135 / 6.876477 (-5.271342)	1.753926 / 2.142072 (-0.388146)	0.659929 / 4.805227 (-4.145298)	0.118093 / 6.500664 (-6.382571)	0.041252 / 0.075469 (-0.034217)

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)	1.009177 / 1.841788 (-0.832610)	11.959624 / 8.074308 (3.885316)	10.484672 / 10.191392 (0.293280)	0.142085 / 0.680424 (-0.538339)	0.015955 / 0.534201 (-0.518245)	0.283649 / 0.579283 (-0.295634)	0.125681 / 0.434364 (-0.308683)	0.320490 / 0.540337 (-0.219847)	0.440353 / 1.386936 (-0.946583)

[albertvillanova]

Maybe a patch release will be needed with this fix.