{"Package":"keyed","Title":"Explicit Key Assumptions for Flat-File Data","Version":"0.1.3","Authors@R":"person(\"Gilles\", \"Colling\", email = \"gilles.colling051@gmail.com\",\nrole = c(\"aut\", \"cre\", \"cph\"),\ncomment = c(ORCID = \"0000-0003-3070-6066\"))","Language":"en-US","Description":"Helps make implicit data assumptions explicit by attaching\nkeys to flat-file data that error when those assumptions are violated.\nDesigned for CSV-first workflows without database infrastructure or version\ncontrol. Provides key definition, assumption checks, join diagnostics, and\noptional drift detection against reference snapshots.","License":"MIT + file LICENSE","URL":"https://github.com/gcol33/keyed, https://gillescolling.com/keyed/","BugReports":"https://github.com/gcol33/keyed/issues","Encoding":"UTF-8","RoxygenNote":"7.3.3","Depends":"R (>= 4.1.0)","Imports":"cli, dplyr (>= 1.0.0), digest, pillar, rlang (>= 1.0.0),\ntibble, vctrs (>= 0.5.0)","Suggests":"joinspy, knitr, rmarkdown, testthat (>= 3.0.0), uuid","VignetteBuilder":"knitr","Config/testthat/edition":"3","NeedsCompilation":"no","Packaged":"2026-02-03 21:34:34 UTC; Gilles Colling","Author":"Gilles Colling [aut, cre, cph] (ORCID:\n)","Maintainer":"Gilles Colling ","Repository":"CRAN","Date/Publication":"2026-02-06 19:50:08 UTC","repoType":"cran","tarballUrl":"ftp://cran.r-project.org/pub/R/src/contrib/keyed_0.1.3.tar.gz","jsonAuthors":[{"name":"Gilles Colling","email":"gilles.colling051@gmail.com","maintainer":true}],"readme":"# keyed\n\n[![CRAN status](https://www.r-pkg.org/badges/version/keyed)](https://CRAN.R-project.org/package=keyed)\n[![CRAN downloads](https://cranlogs.r-pkg.org/badges/grand-total/keyed)](https://cran.r-project.org/package=keyed)\n[![Monthly downloads](https://cranlogs.r-pkg.org/badges/keyed)](https://cran.r-project.org/package=keyed)\n[![R-CMD-check](https://github.com/gcol33/keyed/actions/workflows/R-CMD-check.yaml/badge.svg)](https://github.com/gcol33/keyed/actions/workflows/R-CMD-check.yaml)\n[![Codecov test coverage](https://codecov.io/gh/gcol33/keyed/graph/badge.svg)](https://app.codecov.io/gh/gcol33/keyed)\n[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)\n\n**Primary keys for data frames.**\n\nIn databases, you declare `customer_id` as a primary key and the database enforces uniqueness. With CSV and Excel files, you get no such guarantees - duplicates slip in silently.\n\nkeyed brings database-style protections to R data frames through four features:\n\n| Feature | What it does |\n|---------|--------------|\n| **Keys** | Declare unique columns, enforced through transformations |\n| **Locks** | Assert conditions (no NAs, row counts, coverage) |\n| **UUIDs** | Track row identity through your pipeline |\n| **Commits** | Snapshot data to detect drift |\n\n## Installation\n\n```r\n# install.packages(\"pak\")\npak::pak(\"gcol33/keyed\")\n```\n\n---\n\n## 1. Keys\n\nDeclare which columns must be unique - like a primary key in a database.\n\n```r\nlibrary(keyed)\n\n# Declare the key (errors if not unique)\ncustomers <- read.csv(\"customers.csv\") |>\n key(customer_id)\n\n# Composite keys work too\nsales <- key(sales, region, year)\n```\n\n**Keys follow your data through transformations:**\n\n```r\n# Base R\nactive <- customers[customers$status == \"active\", ]\nhas_key(active)\n#> [1] TRUE\n\n# dplyr\nactive <- customers |> filter(status == \"active\")\nhas_key(active)\n#> [1] TRUE\n```\n\n**Keys block operations that would break uniqueness:**\n\n```r\ncustomers |> mutate(customer_id = 1)\n#> Error: Key is no longer unique after transformation.\n#> i Use `unkey()` first if you intend to break uniqueness.\n\n# To proceed, explicitly remove the key first\ncustomers |> unkey() |> mutate(customer_id = 1)\n```\n\n**Preview joins before running them:**\n\n```r\ndiagnose_join(customers, orders, by = \"customer_id\")\n#> Cardinality: one-to-many\n#> customers: 1000 rows (unique)\n#> orders: 5432 rows (4432 duplicates)\n#> Left join will produce ~5432 rows\n```\n\n---\n\n## 2. Locks\n\nAssert conditions at checkpoints in your pipeline.\n\n```r\ncustomers |>\n lock_unique(customer_id) |> # Must be unique\n lock_no_na(email) |> # No missing emails\n lock_nrow(min = 100) # At least 100 rows\n```\n\nLocks error immediately if the condition fails - no silent continuation.\n\n**Available locks:**\n\n| Function | Checks |\n|----------|--------|\n| `lock_unique(df, col)` | No duplicate values |\n| `lock_no_na(df, col)` | No missing values |\n| `lock_complete(df)` | No NAs in any column |\n| `lock_coverage(df, threshold, col)` | % non-NA above threshold |\n| `lock_nrow(df, min, max)` | Row count in range |\n\n---\n\n## 3. UUIDs\n\nWhen your data has no natural key, generate stable row identifiers.\n\n```r\n# Add a UUID to each row\ncustomers <- add_id(customers)\n#> .id name\n#> 1 a3f2c8e1b9d04567 Alice\n#> 2 7b1e4a9c2f8d3601 Bob\n#> 3 e9c7b2a1d4f80235 Carol\n```\n\n**UUIDs survive all transformations:**\n\n```r\nfiltered <- customers |> filter(name != \"Bob\")\nget_id(filtered)\n#> [1] \"a3f2c8e1b9d04567\" \"e9c7b2a1d4f80235\"\n```\n\n**Track which rows were added or removed:**\n\n```r\ncompare_ids(customers, filtered)\n#> Lost: 1 row (7b1e4a9c2f8d3601)\n#> Kept: 2 rows\n```\n\nUUIDs let you trace rows through joins, filters, and reshaping - essential for debugging data pipelines.\n\n---\n\n## 4. Commits\n\nSnapshot your data to detect unexpected changes later.\n\n```r\n# Save a snapshot (stored in memory for this session)\ncustomers <- customers |> commit_keyed()\n\n# Work with your data...\ncustomers <- customers |>\n filter(status == \"active\") |>\n mutate(score = score + 10)\n\n# Check what changed since the commit\ncheck_drift(customers)\n#> Drift detected!\n#> - Row count: 1000 -> 847 (-153)\n#> - Column 'score' modified\n```\n\n**How it works:**\n- Each data frame can have one snapshot attached\n- Snapshots persist for your R session (lost on restart)\n- `check_drift()` compares current state to the snapshot\n- `clear_snapshot()` removes it, `list_snapshots()` shows all\n\nUseful for catching unexpected changes during interactive analysis.\n\n---\n\n## When to Use Something Else\n\n| Need | Better Tool |\n|------|-------------|\n| Enforced schema | SQLite, DuckDB |\n| Full data validation | pointblank, validate |\n| Production pipelines | targets |\n\nkeyed gives you database-style protections without database infrastructure. For exploratory workflows where SQLite is overkill but silent corruption is unacceptable.\n\n## Documentation\n\n- [Quick Start](https://gillescolling.com/keyed/articles/quickstart.html)\n- [Design Philosophy](https://gillescolling.com/keyed/articles/philosophy.html)\n- [Function Reference](https://gillescolling.com/keyed/reference/index.html)\n\n## License\n\nMIT","jobInfo":{"package":"keyed","version":"0.1.3","parsingStatus":"success","parserVersion":2,"parsedAt":"2026-02-06T21:30:19+0000"}}

Description

Usage

Arguments

Value