Douglass.jl

Douglass.jl is a package for manipulating DataFrames in Julia using a syntax that is very similar to Stata.

Installation

Douglass is not registered. To install, type ] in the Julia command prompt, followed by

add https://github.com/jmboehm/Douglass.jl.git

Examples

using Douglass, RDatasets
df = dataset("datasets", "iris")
# set the active DataFrame
Douglass.set_active_df(:df)

# create a variable `z` that is the sum of `SepalLength` and `SepalWidth`, for each row
d"gen :z = :SepalLength + :SepalWidth"
# replace `z` by the row index for the first 10 observations
d"replace :z = _n if _n <= 10"
# drop a variable
d"drop :z"
# construct the within-group mean for a subset of the observations
d"bysort :Species : egen :z = mean(:SepalLength) if :SepalWidth .> 3.0"

Commands implemented

generate -- Creates a new variable and assigns the output from an expression to it.
replace -- Recplaces the content of a variable, but does not change the type.
egenerate (or egen for short) -- Creates a new variable. Operates on vectors.
ereplace (or erep for short) -- Analogous to egen, replaces values of existing variables.
drop -- Drops the specified observations (if used in conjunction with if) or variables (without if)
rename -- Rename a variable
sort -- Sort the rows activate DataFrame by the specified columns
reshape -- Reshape the activate DataFrame between wide and long format (reshape_long, reshape_wide)
merge -- Merge the active DataFrame with another one in the local scope (merge_m1, merge_1m, merge_11)
duplicates_drop -- Delete duplicate rows, also by subset of columns

See the commands documentation page for more details on syntax of these commands.

REPL mode

Press the backtick (`) to switch between the normal Julia REPL and the Douglass REPL mode:

Benchmarks

These benchmarks are made using a synthetic dataset with 1m observations, on my Macbook Pro (Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz, Julia 1.6.2).

Notes

Better documentation of the interface will come when the package is a bit more stable. In the meantime, the Test script is probably the best introduction to the interface for those that know Stata.
Keep in mind that this is not Stata. Here are some notable differences.

Roadmap / Todo's

Implement more commands
If other people find the package useful, it may be worth making the package extensible, so that other commands can be added in separate packages

Misc

Douglass.jl is named in honour of the economic historian Douglass North.

matthieugomez/Douglass.jl