Documentation
¶
Overview ¶
Package search is full-text search for an app's data and content: in SQLite with FTS5, in PostgreSQL with tsvector and a GIN index, or in memory, for content collections and other data that fits in it.
An index has fields, such as a title and a body, and documents with an ID, a scope and a value for each field:
idx, err := search.New(db, "links", search.Options{
Fields: []search.Field{{Name: "title", Boost: 2}, {Name: "url"}},
})
err = idx.Migrate(ctx)
err = idx.Put(ctx, search.Doc{ID: "42", Scope: "user:7", Fields: map[string]string{"title": "Go 1.26", "url": "https://go.dev"}})
res, err := idx.Search(ctx, search.Query{Text: "go", Scope: "user:7"})
Searches match words as typed, as prefixes of words in the documents, so results come in while someone types, and rank them with BM25 (SQLite and memory) or ts_rank (PostgreSQL). Each hit has highlighted snippets of its fields, as segments of text rather than HTML. Tern UI's SearchBox shows them.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
Types ¶
type Doc ¶
type Doc struct {
// ID identifies the document, such as its row's key or a page's slug.
ID string
// Scope is who the document belongs to, such as "user:7" or "team:3":
// a query with a scope finds only its documents. "" is no one's.
Scope string
// Fields holds the text of each field. Fields not in the index are ignored.
Fields map[string]string
}
Doc is a document in an index.
type Field ¶
type Field struct {
// Name is the field's name, a letter followed by letters, digits and
// underscores.
Name string
// Boost weighs matches in the field: 2 counts them twice. Default 1.
Boost float64
}
Field is a field of an index's documents.
type Hit ¶
type Hit struct {
ID string `json:"id"`
Scope string `json:"scope,omitempty"`
Score float64 `json:"score"`
// Snippets are the fields' text around what matched, with the matches
// marked; long fields are cut to a few words around the first match.
Snippets map[string][]Segment `json:"snippets"`
}
Hit is a document that matched.
type Index ¶
type Index struct {
// contains filtered or unexported fields
}
Index is a search index. It's safe for concurrent use.
func New ¶
New returns an index named name in db, SQLite or PostgreSQL. Call Migrate to create its table.
func NewMemory ¶
NewMemory returns an index in memory, for content collections and other documents that fit in memory and are indexed when the app starts.
type Options ¶
type Options struct {
// Fields are the documents' fields, the most important first.
Fields []Field
// Language is PostgreSQL's text search configuration, for word stems
// such as "running" matching "run": "english" by default, "simple"
// for none. SQLite and memory indexes match words without stems.
Language string
}
Options configure an index.
type Query ¶
type Query struct {
// Text is what was typed. Each word must match, as a word or the
// start of one; quotes and operators are taken as text.
Text string
// Scope limits the search to documents with this scope; "" searches
// documents of every scope.
Scope string
// Limit is how many hits to return, 20 by default; Offset skips some, for pages.
Limit, Offset int
}
Query is a search.
Source Files