module pixeltable.functions.json

Pixeltable UDFs for JsonType. Example:

import pixeltable as pxt
import pixeltable.functions as pxtf

t = pxt.get_table(...)
t.select(pxtf.json.make_list(t.json_col)).collect()

iterator list_iterator()

Signature

@pxt.iterator
list_iterator(
    elements: pxt.Json[(Json, ...)] | None = None,
    *,
    mode: pxt.String = 'strict',
    **kwargs
)

Iterator over elements of a list or lists. There are two distinct call patterns: either a single positional argument; or one or more keyword arguments.

If a single positional argument is specified, as in list_iterator(t.col), then the elements of t.col must contain lists of dictionaries with matching signatures (identical keys and compatible value types). The iterator will yield one new column for each key in the dictionaries, and one output row per element in the lists.
If multiple keyword arguments are specified, as in list_iterator(val_1=t.col_1, val_2=t.col_2), then the elements of each input column must contain lists, but not necessarily lists of dictionaries. The iterator will yield one new column for each keyword argument, zipping together the individual lists.

All of the inputs must be typed Json expressions. Untyped Json will be rejected (the type schema is necessary in order for Pixeltable to determine the types of the output columns). Parameters:

elements (pxt.Json[(Json): A list of dictionaries to iterate over. The dictionary keys will be used as column names in the output. Cannot be specified together with keyword arguments.
mode (Any): Only applies when called with keyword arguments. Determines how to handle lists of different lengths:
- 'strict': Raises an error if the input lists have different lengths.
- 'truncated': Iterates until the shortest input list is exhausted, ignoring any remaining elements in longer lists.
- 'padded': Iterates until the longest input list is exhausted, yielding None for any missing elements from shorter lists.
**kwargs (Any): One or more lists to iterate over. The kwarg names will be used as column names in the output. Cannot be specified together with elements.

uda make_list()

Signature

@pxt.uda
make_list(*args, **kwargs) -> pxt.Json[(Json, ...)]

Collects arguments into a list.

udf concat()

Signature

@pxt.udf
concat(self: pxt.Json | None, other: pxt.Json | None) -> pxt.Json

Concatenate two JSON arrays into a new array. Returns null if either operand is null. Raises if either operand is a non-null, non-array value. Examples:

t.select(t.tags.concat(t.more_tags)).collect()

udf contains()

Signature

@pxt.udf
contains(self: pxt.Json, value: pxt.Json) -> pxt.Bool

Return True if value is an element of a JSON array or a key of a JSON object; False otherwise. Examples:

t.where(t.detections.labels.contains('person')).collect()

udf count()

Signature

@pxt.udf
count(self: pxt.Json, value: pxt.Json) -> pxt.Int

Return the number of times value occurs in a JSON array. Examples:

t.select(t.detections.labels.count('person')).collect()

udf dumps()

Signature

@pxt.udf
dumps(obj: pxt.Json) -> pxt.String

Serialize a JSON object to a string. Equivalent to json.dumps(). Parameters:

obj (pxt.Json): A JSON-serializable object (dict, list, or scalar).

Returns:

pxt.String: A JSON-formatted string.

udf flatten()

Signature

@pxt.udf
flatten(self: pxt.Json) -> pxt.Json

Concatenate the elements of a JSON array one level deep; non-array elements are kept as-is. Examples:

t.select(t.chunks.flatten()).collect()

udf get()

Signature

@pxt.udf
get(
    self: pxt.Json | None,
    key: pxt.String,
    default: pxt.Json | None = None
) -> pxt.Json

Return the value of key if the value is a JSON object containing it, otherwise default. Examples:

t.select(t.metadata.get('author', default='unknown')).collect()

udf is_empty()

Signature

@pxt.udf
is_empty(self: pxt.Json | None) -> pxt.Bool

Return True if the value is null, an empty array, an empty object, or an empty string; False otherwise (including for numbers and booleans). Examples:

t.where(~t.detections.bboxes.is_empty()).collect()

udf items()

Signature

@pxt.udf
items(self: pxt.Json) -> pxt.Json

Return the [key, value] pairs of a JSON object as a list. keys(), values(), and items() share the same ordering. Examples:

t.select(t.metadata.items()).collect()

udf keys()

Signature

@pxt.udf
keys(self: pxt.Json) -> pxt.Json[(String, ...)]

Return the keys of a JSON object. keys(), values(), and items() share the same ordering. Examples:

t.select(t.metadata.keys()).collect()

udf len()

Signature

@pxt.udf
len(self: pxt.Json) -> pxt.Int

Return the number of elements in a JSON array, keys in a JSON object, or characters in a JSON string. Not defined for numbers or booleans. A null value (or missing path) yields null. Examples:

t.select(t.detections.bboxes.len()).collect()

udf max()

Signature

@pxt.udf
max(self: pxt.Json) -> pxt.Float | None

Return the largest number in a JSON array, or null if the array is empty. Examples:

t.select(t.detections.scores.max()).collect()

udf mean()

Signature

@pxt.udf
mean(self: pxt.Json) -> pxt.Float | None

Return the arithmetic mean of the numbers in a JSON array, or null if the array is empty. Examples:

t.select(t.detections.scores.mean()).collect()

udf merge()

Signature

@pxt.udf
merge(self: pxt.Json | None, other: pxt.Json | None) -> pxt.Json

Merge two JSON objects into a new object; on a key conflict the value from other wins. Returns null if either operand is null. Raises if either operand is a non-null, non-object value. Examples:

t.select(t.defaults.merge(t.overrides)).collect()

udf min()

Signature

@pxt.udf
min(self: pxt.Json) -> pxt.Float | None

Return the smallest number in a JSON array, or null if the array is empty. Examples:

t.select(t.detections.scores.min()).collect()

udf sum()

Signature

@pxt.udf
sum(self: pxt.Json) -> pxt.Float

Return the sum of the numbers in a JSON array (0 for an empty array). Examples:

t.select(t.detections.scores.sum()).collect()

udf values()

Signature

@pxt.udf
values(self: pxt.Json) -> pxt.Json

Return the values of a JSON object. keys(), values(), and items() share the same ordering. Examples:

t.select(t.metadata.values()).collect()

​module pixeltable.functions.json

​iterator list_iterator()

​uda make_list()

​udf concat()

​udf contains()

​udf count()

​udf dumps()

​udf flatten()

​udf get()

​udf is_empty()

​udf items()

​udf keys()

​udf len()

​udf max()

​udf mean()

​udf merge()

​udf min()

​udf sum()

​udf values()

module pixeltable.functions.json

iterator list_iterator()

uda make_list()

udf concat()

udf contains()

udf count()

udf dumps()

udf flatten()

udf get()

udf is_empty()

udf items()

udf keys()

udf len()

udf max()

udf mean()

udf merge()

udf min()

udf sum()

udf values()