Skip to content
webinmovePublic

About

A query lib that expose sequelize query abilities

Resources

Stars

0 stars

Watchers

0 watching

Forks

Latest commit

 

History

10 Commits

Folders and files

Repository files navigation

Kuuliza

Description

Kuuliza means "query" in Swahili.

This package exposes Sequelize querying abilities from a simple params object. It is designed to be and stay simple !!!

Here is an example:

QueryBuilder.build({
  filters: {
    assignations: { is: null },
    companyIdentifier: { not: null },
    alertShortId: 'ALERT_SHORT_ID',
    eventShortId: [
      'EVENT_SHORT_ID_1',
      'EVENT_SHORT_ID_2',
      'EVENT_SHORT_ID_3'
    ],
    indexedMeta: {
      eventType: 'EVENT_TYPE_VALUE',
      eventSeverity: ['LOW', 'HIGH']
    },
    updatedAt: {
      gte: Date.now()
    }
  },
  from: 0,
  size: 50,
  sort: [{ updatedAt: 'ASC' }]
});

Usage

Initialize the builder once with Sequelize, then call build() (and buildAggregation() / buildSearch()) anywhere.

import Sequelize from 'sequelize';
import QueryBuilder from '@webinmove/kuuliza';

QueryBuilder.init(Sequelize); // pass the whole namespace — enables aggregation & search

const query = QueryBuilder.build({
  filters: {
    status: 'active',
    createdAt: { gte: Date.now() }
  },
  sort: [{ createdAt: 'DESC' }],
  from: 0,
  size: 20
});

const results = await MyModel.findAll(query);

QueryBuilder.init(Op) still works if you only use build(). Aggregations and faceted search need the Sequelize helpers (fn, col, literal, where), so pass the whole namespace — init(Sequelize) — or the legacy init(Op, { fn, col, literal, where }).

Build params

Field Type Required Description
filters object no Key/value pairs mapped to Sequelize where conditions
sort array no Array of { field: 'ASC' | 'DESC' } objects
from number no Offset (default: 0)
size number no Limit (default: 100)
attributes string[] no Columns to select

Filter operators

Filters support the following operators as object keys:

  • eq ne — equality / inequality
  • gt gte lt lte — comparisons
  • in notIn — array membership
  • like notLike iLike notILike — pattern matching
  • between notBetween — range
  • overlap — array overlap
  • is not — null checks (use { is: null } or { not: null })
  • or — OR condition. Takes an array of values/predicates, e.g. { or: [60, 130] } or { or: [{ gt: 100 }, { lt: 10 }] }. A single nested operator object ({ or: { gt: 1 } }) is also accepted.

Nested objects (e.g. JSON columns) are supported — each key in the nested object is treated as a sub-filter.

Output

build() returns a plain object ready to pass to any Sequelize finder:

{
  attributes: undefined,    // or string[]
  where: { ... },           // Sequelize WhereOptions
  order: [['field', 'ASC']] // or null
  limit: 100,
  offset: 0
}

Behavior note: filter values '', 0 and false are real equality conditions — only null/undefined mean IS NULL.

Aggregations

buildAggregation() turns a params object into a grouped findAll options object (raw: true, subQuery: false). It reuses filters, sort, from and size, and leaves build() untouched.

const options = QueryBuilder.buildAggregation({
  filters: { archived: false },
  groupBy: { field: 'createdAt', interval: 'month' }, // 'field' | { field, interval } | [ …mix ]
  metrics: {
    count: 'count',                       // shorthand → COUNT(*)
    revenue: { fn: 'sum', field: 'amount' },
    avgAmount: { fn: 'avg', field: 'amount' }
  },
  having: { count: { gte: 10 } },         // filter buckets — keyed by metric alias
  sort: [{ count: 'DESC' }]               // order by a metric alias or a groupBy field
});

const rows = await MyModel.findAll(options);
// rows: [ { createdAt: <Date>, count: '42', revenue: '15000.50', avgAmount: '357.15' }, … ]
Field Description
groupBy A field name, a date bucket { field, interval }, or an array of either. Omit it for global stats (a single row).
metrics Object of alias → spec. Each alias becomes the output column. Spec is 'count' or { fn, field }.
having Object of alias → { operator: value }, e.g. { count: { gte: 10 } }.
sort [{ field-or-alias: 'ASC'/'DESC' }]. Sorting by a metric alias gives you top‑N buckets.
from / size Offset / limit over groups (rejected when there is no groupBy).
  • Aggregate functions: count, countDistinct, sum, avg, min, max. count with no field is COUNT(*); the others require a field.

  • Date intervals: hour, day, week, month, quarter, year. The date-bucket form takes an optional timezone ({ field, interval, timezone }, default 'UTC') so DATE_TRUNC boundaries are deterministic regardless of the DB session timezone — requires PostgreSQL 14+.

  • Numbers come back as strings (Postgres returns aggregates as strings under raw). Pass them through coerceAggregation() to turn count/sum/avg into JS numbers (min/max are left untouched — they are type‑ambiguous):

    const rows = QueryBuilder.coerceAggregation(await MyModel.findAll(options), params);
  • size limits the number of groups returned, not the query cost — it is not DoS protection.

fieldMap — snake_case / underscored models

kuuliza wraps metric targets and date-bucket fields in col(), which emits the name verbatim — it does not resolve a Sequelize model's field: mapping. If your model is underscored (or otherwise maps camelCase attributes to snake_case columns), pass fieldMap: { attribute: dbColumn } so those col() identifiers target the real column:

// Model: { distributionRequestShortId: { field: 'distribution_request_short_id' }, ... }
const fieldMap = Object.fromEntries(
  Object.entries(Model.rawAttributes).map(([attr, def]) => [attr, def.field || attr])
);

QueryBuilder.buildAggregation({
  groupBy: { field: 'createdAt', interval: 'day' },             // bucket on created_at
  metrics: { uniq: { fn: 'countDistinct', field: 'titleId' } }, // COUNT(DISTINCT title_id)
  fieldMap                                                      // { createdAt:'created_at', titleId:'title_id', … }
});
// date-bucket rows still come back keyed by the attribute name (`createdAt`), not the column.
  • Plain groupBy dimensions need no entry — they are emitted as bare attribute strings, which Sequelize maps to columns (GROUP BY then references the output alias).
  • Only metric fields ({ fn, field }), date-bucket fields, and facet fields (terms value/group, stats min/max/avg) are resolved through fieldMap.
  • filters and disjunctive-facet drops always key on the attribute name (Sequelize maps those), so they ignore fieldMap. buildSearch() accepts the same fieldMap.

Faceted search

buildSearch() produces the option specs for a search with facet counts — a hits query, a total‑count spec, and one query per facet — and returns them without executing. The controller runs them in parallel and assembles the response with assembleSearch().

Disjunctive facets compute their counts with their own filter removed (so a selected facet still shows the other values' counts), while every other facet keeps it.

const spec = QueryBuilder.buildSearch({
  filters: { brand: 'Nike', size: '42' },
  from: 0,
  size: 20,
  facets: {
    brand: { type: 'terms', disjunctive: true, limit: 10 },
    size:  { type: 'terms', disjunctive: true },
    color: { type: 'terms', disjunctive: true },
    price: { type: 'stats' }              // min / max / avg / count
  }
});

const [hits, total, ...facetRows] = await Promise.all([
  MyModel.findAll(spec.hits),
  MyModel.count(spec.count),
  ...spec.facets.map((f) => MyModel.findAll(f.options))
]);

const response = QueryBuilder.assembleSearch({
  hits,
  total,
  facetResults: spec.facets.map((f, i) => ({ name: f.name, type: f.type, rows: facetRows[i] }))
});
  • Facet types: terms (value/count buckets, disjunctive defaults to true, limit defaults to 10) and stats (numeric min/max/avg/count, disjunctive defaults to false).
  • Field validation is the consuming app's responsibility. kuuliza is schema‑agnostic; it never executes a query, and field names are identifier‑escaped via col(). Restrict which fields are facetable in your controller/validator.

Response envelope

The builder returns Sequelize options; your controller assembles the HTTP response. The recommended envelope is stable and additive — every field is typed once and appears only when the request asks for it; a field never changes type:

{
  hits:    [ … ],                          // list / search rows           (array)
  total:   123,                            // matching count               (number)
  buckets: [ { …dimensions, …metrics } ],  // aggregation rows             (array)
  facets:  { brand: [ { value, count } ] },// facet distributions          (object)
  stats:   { price: { min, max, avg, count } } // numeric facet stats      (object)
}

About

A query lib that expose sequelize query abilities

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages