Issue found on page 'Loading JSON'

[maxmoehl]

I was importing large JSON objects with many fields and duckdb refused to expand each field into its own column and instead give me a single STRUCT... or MAP(VARCHAR, VARCHAR) column. When messing around with the parameters I stumbled across this list of available arguments:

D SELECT * FROM read_json('./test-flat.json', format='array', auto_detectt=true) LIMIT 1;
Binder Error: Invalid named parameter "auto_detectt" for function read_json
Candidates:
    map_inference_threshold BIGINT
    records VARCHAR
    timestampformat VARCHAR
    field_appearance_threshold DOUBLE
    date_format VARCHAR
    dateformat VARCHAR
    sample_size BIGINT
    columns ANY
    convert_strings_to_integers BOOLEAN
    format VARCHAR
    ignore_errors BOOLEAN
    maximum_object_size UINTEGER
    maximum_depth BIGINT
    auto_detect BOOLEAN
    union_by_name BOOLEAN
    maximum_sample_files BIGINT
    compression VARCHAR
    timestamp_format VARCHAR
    hive_types ANY
    hive_partitioning BOOLEAN
    hive_types_autocast BOOLEAN
    filename ANY

It seems like not all of them are documented, the one I needed in my case was map_inference_threshold which apparently has a default which is too low for my objects, after setting it to -1 the columns got generated. Searching via GitHub also yields no results.

Page URL: https://duckdb.org/docs/data/json/loading_json.html

[szarnyasg]

Hi @maxmoehl, thanks for reporting this. Indeed there are missing flags – I'm working on documenting them in #3922.