LimeSurvey API - Master branch

ResponseAggregateBatch

in package

• Application

Collects every aggregate needed for survey statistics and executes them as conditional aggregates over a single scan of the responses table.

Table of Contents

Constants

MAX_EXPRESSIONS_PER_QUERY  = 1500

Hard cap of SELECT expressions per query to bound per-row CASE cost

FIELD_SEPARATOR  = "\x1e"

Separator joining the columns of a multi-field aggregate into one field key

KIND_ANY_NON_EMPTY  = 'anyNonEmpty'

KIND_BLANK  = 'blank'

KIND_JSON_ELEMENT  = 'jsonElement'

KIND_MAX  = 'max'

KIND_MIN  = 'min'

KIND_NON_EMPTY  = 'nonEmpty'

KIND_NUMERIC  = 'numeric'

KIND_SUM  = 'sum'

KIND_SUM_SQUARES  = 'sumSquares'

KIND_TOTAL  = 'total'

KIND_VALUE  = 'value'

MAX_MEDIANS_PER_QUERY  = 100

NUMERIC_PATTERN  = '^-?[0-9]*\\.?[0-9]+$'

Plain decimal test for numeric answers stored in text columns

Properties

$aliasIndex  : array<string, string>

$executed  : bool

$filters  : StatisticsResponseFilters|null

$medianRequests  : array<string, array{field: string, countAlias: string}>

Median cannot be expressed as a one-scan conditional aggregate, so these run as one ordered LIMIT/OFFSET query per field after the main pass.

$requests  : array<string, array{kind: string, field: string, value: string}>

$results  : array<string, int|float>

$surveyId  : int

Methods

__construct()  : mixed

countAnyNonEmpty()  : string

Count of rows where at least one of the given columns is neither NULL nor empty

countBlank()  : string

Count of rows where the column is NULL or empty.

countJsonArrayValue()  : string

Count of rows where the JSON-array column holds the given value at the given zero-based position (rankings store the whole answer as one JSON array of codes, the element index being the rank).

countNonEmpty()  : string

Count of rows where the column is neither NULL nor empty.

countNumeric()  : string

Count of rows whose cell holds a numeric value. For numeric columns this equals countNonEmpty(); for text columns non-numeric cells are excluded.

countTotal()  : string

Total row count (with the response filters applied).

countValue()  : string

Count of rows where the column equals the given value.

execute()  : void

Execute all registered aggregates over the responses table.

isExecuted()  : bool

maxValue()  : string

Largest numeric value in a column (empty cells ignored).

medianValue()  : string

minValue()  : string

Smallest numeric value in a column (empty cells ignored).

sumSquares()  : string

Sum of the squared numeric values in a column (non-numeric/empty cells count as 0). Combined with sumValues() and countNonEmpty() this yields the population variance/standard deviation without a second scan.

sumValues()  : string

Sum of the numeric values in a column (non-numeric/empty cells count as 0). Combined with countNonEmpty() this yields a column mean. The result preserves up to 4 decimal places, so fractional answers (e.g. 1.5) are not lost.

value()  : int|float

Resolved value for a previously registered alias (0 before execute()).

buildExpression()  : string

buildMedianSelect()  : string

buildSearchConditions()  : array<string|int, string>

One condition group per search term: the term must appear (case-insensitively) in at least one free-text answer column; the terms themselves combine with AND. A survey without any text answer column cannot match a term, so the whole filter becomes FALSE.

buildWhere()  : string

executeMedians()  : void

Median via an ordered LIMIT/OFFSET sub-select: the middle value, or the mean of the two middle values for even counts. There is no portable single-scan SQL median, but the per-field sub-selects are merged into chunked UNION ALL statements so resolving N fields (e.g. every cell of an array question) does not cost N round trips.

getDb()  : CDbConnection

getSearchableColumns()  : array<string|int, string>

isJsonColumn()  : bool

jsonArrayElementCheck()  : string

Predicate testing the element at a zero-based position of a JSON-array column against a value. Extracting in SQL keeps per-rank counts inside the shared single scan instead of pulling every row into PHP. The merged query must survive cells that are not valid JSON (encrypted rankings live in plain text columns): MySQL/MSSQL guard per row with JSON_VALID()/ISJSON(); Postgres only gets the ->> operator when the column really is json/jsonb, because on a text column the operator itself is a hard SQL error.

nonEmptyCheck()  : string

Answered-cell predicate for a column. Numeric response columns (e.g.

numericCellCheck()  : string

Castable-cell predicate guarding every CAST(... AS DECIMAL). Numeric answers of array questions live in text columns which may hold non-numeric junk (API-submitted or legacy data); casting that fails the whole query on Postgres and coerces to 0 (skewing sums) on MySQL, so text columns get a per-driver numeric test.

register()  : string