View Source Paradex (Paradex v0.2.0)
A collection of macros for composing ParadeDB queries.
Summary
Macros
Returns the BM25 search score for each result, aliasing it as pdb_score in the query. Can be used with score/0 to order results
Same as score/1 but uses a variable for the alias instead. Does not support literals.
Macro for paradedb.snippet, used for highlighting.
Macro for the @@@ full text search operator.
Term-level queries
Macro for paradedb.fuzzy_term.
Macro for paradedb.exists.
Macro for paradedb.range_term.
Macro for paradedb.regex. Be mindful that these regular expressions follow Tantivy's syntax, which follows Rust's regex crate with a few variations.
Macro for paradedb.term.
Macro for paradedb.term_set.
Range queries
Macro for paradedb.range using the date data type.
Macro for paradedb.range using the int4 data type.
Macro for paradedb.range using the int8 data type.
Macro for paradedb.range using the timestamp data type.
Compound queries
Macro for paradedb.all.
Macro for paradedb.boolean.
Macro for paradedb.boost.
Macro for paradedb.const_score.
Macro for paradedb.empty.
Macros
Returns the BM25 search score for each result, aliasing it as pdb_score in the query. Can be used with score/0 to order results:
from(
c in Call,
select: {c, score(c.id)},
where: c.transcript ~> "mechanic",
order_by: [desc: score()]
)Same as score/1 but uses a variable for the alias instead. Does not support literals.
score_alias = "my_score"
from(
c in Call,
select: {c, score_as(c.id, ^score_alias)},
where: c.transcript ~> "mechanic",
order_by: [desc: score_as(^score_alias)]
)snippet(field, start_tag \\ "<b>", end_tag \\ "</b>", max_num_chars \\ 150)
View Source (macro)Macro for paradedb.snippet, used for highlighting.
from(
c in Call,
select: {c, snippet(c.transcript)},
where: c.transcript ~> "mechanic"
)Macro for the @@@ full text search operator.
~> is used as it's one of a few infix operators Elixir's capable of parsing, but aren't presently used.
Examples
Search queries can be run on fields directly:
import Paradex
from(
c in Call,
where: c.transcript ~> "bus"
)Alternatively a key field and query object can be used for advanced queries:
from(
c in Call,
where: c.id ~> disjunction_max([
parse("transcript:bus"),
int4range("call_length", 10, nil, "[)")
])
)Term-level queries
fuzzy_term(field, value, distance \\ 2, transpose_cost_one \\ true, prefix \\ false)
View Source (macro)Macro for paradedb.fuzzy_term.
from(
c in Call,
where: c.id ~> fuzzy_term("transcript", "bus", 2, true, false)
)Macro for paradedb.exists.
Prefixed with pdb_ to avoid conflicting with Ecto.Query.API.exists/1.
from(
c in Call,
where: c.id ~> pdb_exists("call_length")
)Macro for paradedb.range_term.
Macro for paradedb.regex. Be mindful that these regular expressions follow Tantivy's syntax, which follows Rust's regex crate with a few variations.
from(
c in Call,
where: c.id ~> regex("transcript", "(stop|route)")
)Macro for paradedb.term.
from(
c in Call,
where: c.id ~> term("talkgroup_num", 7695)
)Macro for paradedb.term_set.
from(
c in Call,
where: c.id ~> term_set([
term("talkgroup_num", 7700),
term("call_length", 20)
])
)Range queries
Macro for paradedb.range using the date data type.
start = ~D[2024-10-09]
stop = ~D[2024-10-10]
query =
from(
c in Call,
where: c.id ~> daterange("start_time", ^start, ^stop, "[]")
)Macro for paradedb.range using the int4 data type.
from(
c in Call,
where: c.id ~> int4range("call_length", 5, nil, "[)")
)Macro for paradedb.range using the int8 data type.
from(
c in Call,
where: c.id ~> int8range("call_length", 5, nil, "[)")
)Macro for paradedb.range using the timestamp data type.
begin = ~U[2024-10-09 08:00:00.00Z]
query =
from(
c in Call,
where: c.id ~> tsrange("start_time", ^begin, nil, "[)")
)Phrase-level queries
fuzzy_phrase(field, value, distance \\ 2, transpose_cost_one \\ true, prefix \\ false, match_all_terms \\ false)
View Source (macro)Macro for paradedb.fuzzy_phrase.
from(
c in Call,
where: c.id ~> fuzzy_phrase("transcript", "bus sotp")
)Macro for paradedb.phrase.
from(
c in Call,
where: c.id ~> phrase("transcript", ["bus", "stop"], 1)
)Macro for paradedb.phrase_prefix.
from(
c in Call,
where: c.id ~> phrase_prefix("transcript", ["en"])
)Compound queries
Macro for paradedb.all.
from(
c in Call,
where: c.id ~> all()
)Macro for paradedb.boolean.
Each value must be a literal Keyword list at the top level, where each value is a list:
x = "transcript:transfer"
from(
c in Call,
where: c.id ~> boolean(
must: [parse(^x)],
must_not: [parse("transcript:station")]
)
)Keys other than must, should, and must_not are ignored, so be wary of misspelling.
Macro for paradedb.boost.
from(
c in Call,
select: {c. score(c.id)},
boost(2.0, "transcript:bus")
)Macro for paradedb.const_score.
from(
c in Call,
select: {c, score(c.id)},
where: c.id ~> const_score(2.0, parse("transcript:bus"))
)Macro for paradedb.disjunction_max.
from(
c in Call,
where:
c.id ~> disjunction_max([
parse("transcript:bus"),
int4range("call_length", 10, nil, "[)")
])
)Macro for paradedb.empty.
from(
c in Call,
where: c.id ~> empty()
)Macro for paradedb.parse.
from(
c in Call,
where: c.id ~> parse("transcript:bus")
)parse_with_field(field, query, lenient \\ false, conjunction_mode \\ true)
View Source (macro)Macro for paradedb.parse_with_field.
from(
c in Call,
where: c.id ~> parse_with_field("transcript", "traffic congestion")
)