DBDelve
DBDelve는 Rust 기반 빠른 다중 엔진 데이터베이스 클라이언트, 무료 오픈소스, 텔레메트리 없음
연구 데이터셋의 데이터 유형을 판별하고 어휘를 제공하며, 식별자·지리·시간 형식·통화·코드·해시 등을 탐지하는 라이브러리를 포함한다
A vocabulary and detection engine for the data types found in research datasets. Provides a library of column detectors (identifiers, geographies, temporal formats, currencies, codes, hashes, and more
개발 도구RGitHub @lecy온라인운영자 미확인
써 보셨어요? 한 번만 눌러 주세요. 로그인은 필요 없어요.
formex: FORMat EXpressions for detecting and standardizing semantic data types
README 전체를 불러오는 중입니다.
README 전문 보기 ↗formex (format expressions) is a vocabulary + detection + transformation engine for the column types found in research data. Point it at a column and it tells you what the values are — a calendar date, an ISBN, a US state, a person name, a currency amount — and then standardizes them to a stable, analysis-ready form.
It is built around a versioned ontology of semantic types, a library of deterministic detectors, and a two-stage router → verifier classifier that returns a graceful, confidence-aware answer rather than a forced guess.
# install.packages("remotes")
remotes::install_github("lecy/formex")
The learned router additionally uses xgboost (a soft dependency); without it,
formex falls back to a model-free signature router.
guess_data_type() — what is this column?library(formex)
guess_data_type(c("03/15/2019", "12/31/2020", "01/01/2021"))$route
#> data_type "temporal"
#> semantic_family "date"
#> semantic_type "calendar"
#> format_label "calendar"
#> depth 4 # resolved all the way to a leaf
#> confidence 1.00
The $route record anchors the column to the ontology
(data_type → semantic_family → semantic_type → format_label) and truncates
gracefully: when the values cannot pin a leaf, it stops at the deepest level it
can actually defend (depth 0–4). The legacy detector-based fields
($guess, $ontology) are returned alongside, unchanged.
fx_stabilize() — put it in a standard formfx_stabilize(c("isbn 0395629764", "ISBN 957-33-0471-6"),
"administrative", "isbn")$stable
#> "0395629764" "9573304716"
fx_stabilize(c("California", "TX", "Ohio"), "geographic", "us_state")$stable
#> "CA" "TX" "OH"
Each ontology row carries a recipe (a small transform DSL) that normalizes the raw values and reports the stable format + import rule.
fx_extract_features()), and a
model-free signature router. guess_data_type(router=) picks the strategy.fx_ontology()) is the shared vocabulary — stable
semantic_type_ids and per-type recipes — that guess_data_type() and
fx_stabilize() both speak.The shipped person/place gazetteers are public-domain: the top 1000 SSA given
names (CC0, via babynames) and the top 2000 US Census 2010 surnames. See
data-raw/gazetteers.R to regenerate them.