SET allow_experimental_object_type = 1;

CREATE TABLE json_test
(
    `raw` JSON
)
ENGINE = MergeTree
ORDER BY tuple()
SETTINGS index_granularity = 8192;

## single row - ok
INSERT INTO json_test SELECT format('{{"bad_subkey":{{"key{}":[{{"value":"{}","bad":true}}]}}}}', toString(number), toString(rand())) FROM numbers(1)

select * from json_test format JSONEachRow settings output_format_json_named_tuples_as_objects = 1;

{"raw":{"bad_subkey":{"key0":[{"bad":1,"value":"3736142820"}]}}}

## up to 1140 rows - slow, but still kinda works
INSERT INTO json_test SELECT format('{{"bad_subkey":{{"key{}":[{{"value":"{}","bad":true}}]}}}}', toString(number), toString(rand())) FROM numbers(1140)

-- MemoryTracker: Peak memory usage (for query): 72.13 MiB.
--- TCPHandler: Processed in 0.797080049 sec.

## from 1141 rows - memory usage suddenly jumps to gigabytes

INSERT INTO json_test SELECT format('{{"bad_subkey":{{"key{}":[{{"value":"{}","bad":true}}]}}}}', toString(number), toString(rand())) FROM numbers(1141)

--- executeQuery: Read 1141 rows, 8.91 KiB in 1.175646399 sec., 970 rows/sec., 7.58 KiB/sec.
---  MemoryTracker: Peak memory usage (for query): 6.78 GiB.

# With more than few thousands rows - unusable completely