krangl
is a {K}otlin library for data w{rangl}ing. By implementing a grammar of data manipulation using a modern functional-style API, it allows to filter, transform, aggregate and reshape tabular data.
krangl
is heavily inspired by the amazing dplyr
for R. krangl
is written in Kotlin, excels in Kotlin, but emphasizes as well on good java-interop. It is mimicking the API of dplyr
, while carefully adding more typed constructs where possible.
If you're not sure about how to proceed, check out krangl in 10 minutes section in the krangl user guide.
To get started simply add it as a dependency via Jcenter:
implementation "de.mpicbg.scicomp:krangl:0.13"
You can also use JitPack with Maven or Gradle to build the latest snapshot as a dependency in your project.
repositories {
maven { url 'https://jitpack.io' }
}
dependencies {
implementation 'com.github.holgerbrandl:krangl:-SNAPSHOT'
}
To build and install it into your local maven cache, simply clone the repo and run
./gradlew install
-
Filter, transform, aggregate and reshape tabular data
-
Modern, user-friendly and easy-to-learn data-science API
-
Reads from plain and compressed tsv, csv, json, or any delimited format with or without header from local or remote
-
Supports grouped operations
-
Ships with JDBC support
-
Tables can contain atomic columns (int, double, boolean) as well as object columns
-
Reshape tables from wide to long and back
-
Table joins (left, right, semi, inner, outer)
-
Cross tabulation
-
Descriptive statistics (mean, min, max, median, ...)
-
many more...
krangl
is just about data wrangling. For data visualization we recommend kravis
which seamlessly integrates with krangl and implements a grammar to build a wide variety of plots.
// Read data-frame from disk
val iris = DataFrame.readTSV("data/iris.txt")
// Create data-frame in memory
val df: DataFrame = dataFrameOf(
"first_name", "last_name", "age", "weight")(
"Max", "Doe", 23, 55,
"Franz", "Smith", 23, 88,
"Horst", "Keanes", 12, 82
)
// Or from csv
// val otherDF = DataFrame.readCSV("path/to/file")
// Print rows
df // with implict string conversion using default options
df.print(colNames = false) // with custom printing options
// Print structure
df.schema()
// Add columns with mutate
// by adding constant values as new column
df.addColumn("salary_category") { 3 }
// by doing basic column arithmetics
df.addColumn("age_3y_later") { it["age"] + 3 }
// Note: krangl dataframes are immutable so we need to (re)assign results to preserve changes.
val newDF = df.addColumn("full_name") { it["first_name"] + " " + it["last_name"] }
// Also feel free to mix types here since krangl overloads arithmetic operators like + for dataframe-columns
df.addColumn("user_id") { it["last_name"] + "_id" + rowNumber }
// Create new attributes with string operations like matching, splitting or extraction.
df.addColumn("with_anz") { it["first_name"].asStrings().map { it!!.contains("anz") } }
// Note: krangl is using 'null' as missing value, and provides convenience methods to process non-NA bits
df.addColumn("first_name_initial") { it["first_name"].map<String>{ it.first() } }
// or add multiple columns at once
df.addColumns(
"age_plus3" to { it["age"] + 3 },
"initials" to { it["first_name"].map<String> { it.first() } concat it["last_name"].map<String> { it.first() } }
)
// Sort your data with sortedBy
df.sortedBy("age")
// and add secondary sorting attributes as varargs
df.sortedBy("age", "weight")
df.sortedByDescending("age")
df.sortedBy { it["weight"].asInts() }
// Subset columns with select
df.select2 { it is IntCol } // functional style column selection
df.select("last_name", "weight") // positive selection
df.remove("weight", "age") // negative selection
df.select({ endsWith("name") }) // selector mini-language
// Subset rows with vectorized filter
df.filter { it["age"] eq 23 }
df.filter { it["weight"] gt 50 }
df.filter({ it["last_name"].isMatching { startsWith("Do") }})
// In case vectorized operations are not possible or available we can also filter tables by row
// which allows for scalar operators
df.filterByRow { it["age"] as Int > 5 }
df.filterByRow { (it["age"] as Int).rem(10) == 0 } // round birthdays :-)
// Summarize
// do simple cross tabulations
df.count("age", "last_name")
// ... or calculate single summary statistic
df.summarize("mean_age" to { it["age"].mean(true) })
// ... or multiple summary statistics
df.summarize(
"min_age" to { it["age"].min() },
"max_age" to { it["age"].max() }
)
// for sake of r and python adoptability you can also use `=` here
df.summarize(
"min_age" `=` { it["age"].min() },
"max_age" `=` { it["age"].max() }
)
// Grouped operations
val groupedDf: DataFrame = df.groupBy("age") // or provide multiple grouping attributes with varargs
val sumDF = groupedDf.summarize(
"mean_weight" to { it["weight"].mean(removeNA = true) },
"num_persons" to { nrow }
)
// Optionally ungroup the data
sumDF.ungroup().print()
// generate object bindings for kotlin.
// Unfortunately the syntax is a bit odd since we can not access the variable name by reflection
sumDF.printDataClassSchema("Person")
// This will generate and print the following conversion code:
data class Person(val age: Int, val mean_weight: Double, val num_persons: Int)
val records = sumDF.rows.map { row -> Person(row["age"] as Int, row["mean_weight"] as Double, row["num_persons"] as Int) }
// Now we can use the krangl result table in a strongly typed way
records.first().mean_weight
// Vice versa we can also convert an existing set of objects into
val recordsDF = records.asDataFrame()
recordsDF.print()
// to populate a data-frame with selected properties only, we can do
val deparsedDF = records.deparseRecords { mapOf("age" to it.age, "weight" to it.mean_weight) }
krangl
is not yet mature, full of bugs and its API is in constant flux. Nevertheless, feel welcome to submit pull-requests or tickets, or simply get in touch via gitter (see button on top).
- Krangl User Guide for detailed information about the API and usage examples.
- API Docs for detailed information about the API including manu usage examples
- TBD
krangl
Cheat Sheet
Another great introduction into data-science with kotlin was presented at 2019's KotlinConf by Roman Belov from JetBrains.
Feel welcome to post ideas, suggestions and criticism to our tracker.
We always welcome pull requests. :-)
You could also show your spiritual support by upvoting krangl
here on github.
Also see
- Developer Information with technical notes & details about to build, test, release and improve
krangl
- Roadmap complementing the tracker with a backlog
Also, there are a few issues in the IDE itself which limit the applicability/usability of krangl
, So, you may want to vote for
- KT-24789 "Unresolved reference" when running a script which is a symlink to a script outside of source roots
- KT-12583 IDE REPL should run in project root directory
- KT-11409 Allow to "Send Selection To Kotlin Console"
- KT-13319 Support ":paste" for pasting multi-line expressions in REPL
- KT-21224 REPL output is not aligned with input
Similar APIs (not just Kotlin)
- Scala DataTable: a lightweight, in-memory table structure written in Scala
- joinery implements data frames for Java
- tablesaw which is (according to its authors) the The simplest way to slice data in Java
- paleo which provides immutable Java 8 data frames with typed columns
- agate isa Python data analysis library that is optimized for humans instead of machines
- pandas provides high-performance, easy-to-use data structures and data analysis tools for python (cheatsheet)
- dplyr which is a grammar of data manipulation (R-lang)
- morpheus-core which is a data science framework implementing an R-like data-frame for the JVM
- Frameless is a Scala library for working with Spark using more expressive types, including a more strongly typed Dataset/DataFrame API
Other data-science projects:
- vectorz is a fast and flexible numerical library for Java featuring N-dimensional arrays
- koma is a scientific computing library written in Kotlin, designed to allow development of cross-platform numerical applications
- termsql converts text from a file or from stdin into SQL table and query it instantly. Uses sqlite as backend.
- kotliquery is a handy database access library
- Dex : The Data Explorer is a data visualization tool written capable of powerful ETL and publishing web visualizations
Data Visualization
- kravis which implements a Kotlin DSL for scientific data visualization
- XChart is a light weight Java library for plotting data
- Charts in TornadoFX provide visualzation examples using javaFX