Data Operations¶

These functions extract, filter, transform, and aggregate data.

get¶

get(data, field)

Extracts a single field from a record or structured data object.

Parameters:

data - A record or structured data object
field - A string specifying the field name

Returns: The value of the specified field.

let response = query_model(my_model, "Is AI safe?", {"topic": "safety"});
let answer = get(response, "value");

get_all¶

get_all(data, field)

Extracts a field from each element in a list of records.

Parameters:

data - A list of records
field - A string specifying the field name to extract

Returns: A list of values, one from each record.

let responses = [
    {"value": "yes", "confidence": 0.9},
    {"value": "no", "confidence": 0.6},
    {"value": "yes", "confidence": 0.8}
];
let values = get_all(responses, "value");
// values = ["yes", "no", "yes"]

filter_by_key¶

filter_by_key(data, key, value)

Filters a list of records, keeping only those where the specified field matches the given value.

Parameters:

data - A list of records
key - A string specifying the field name to filter on
value - The value to match against

Returns: A filtered list of records.

let high_confidence = filter_by_key(responses, "confidence", 0.9);

flatten¶

flatten(array)

Flattens a multi-dimensional array into a one-dimensional array.

Parameters:

array - An array (possibly multi-dimensional)

Returns: A flat one-dimensional array.

let flat = flatten([[1, 2], [3, 4]]);
// flat = [1, 2, 3, 4]

argmax¶

argmax(array)

Returns the index of the maximum value for each element in the array. For multi-class probability arrays, this returns the predicted class label.

Parameters:

array - An array of values or an array of probability arrays

Returns: An array of indices.

let predictions = argmax(probability_outputs);

convert_to_one_hot¶

convert_to_one_hot(array)

Converts an array of probability distributions to one-hot encoded vectors.

Parameters:

array - An array of probability arrays

Returns: An array of one-hot vectors where only the maximum probability index is 1.

let one_hot = convert_to_one_hot(probabilities);

fill¶

fill(value, length)

Creates an array of a specified length filled with a single value.

Parameters:

value - The value to repeat
length - The number of elements

Returns: An array of the given length.

let zeros = fill(0, 10);
// zeros = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0]

sum¶

sum(list)

Computes the sum of all elements in a list.

Parameters:

list - A list of numbers

Returns: The sum.

let total = sum([1, 2, 3, 4, 5]);
// total = 15

size¶

size(list)

Returns the number of elements in a list.

Parameters:

list - A list

Returns: The length as an integer.

let n = size([10, 20, 30]);
// n = 3

histogram¶

histogram(data, question_field, answer_field)

Computes a frequency distribution of answers grouped by question.

Parameters:

data - A list of records
question_field - The field name containing the question identifier
answer_field - The field name containing the answer value

Returns: A record mapping questions to answer frequency distributions.

let dist = histogram(responses, "question", "answer");

box_plot¶

box_plot(data, question_field, answer_field, score_field)

Computes box plot statistics grouped by question and answer.

Parameters:

data - A list of records
question_field - The field name for question grouping
answer_field - The field name for answer grouping
score_field - The field name containing numeric scores

Returns: A list of records with box plot statistics (min, max, median, quartiles) for each group.

let plots = box_plot(results, "question", "answer", "score");