Dataset (data frame) manipulation API for the tech.ml.dataset library
[scicloj/tablecloth "4.04"]
tech.ml.dataset is a great and fast library which brings columnar dataset to the Clojure. Chris Nuernberger has been working on this library for last year as a part of bigger tech.ml
stack.
I've started to test the library and help to fix uncovered bugs. My main goal was to compare functionalities with the other standards from other platforms. I focused on R solutions: dplyr, tidyr and data.table.
During conversions of the examples I've come up how to reorganized existing tech.ml.dataset
functions into simple to use API. The main goals were:
- Focus on dataset manipulation functionality, leaving other parts of
tech.ml
like pipelines, datatypes, readers, ML, etc. - Single entry point for common operations - one function dispatching on given arguments.
group-by
results with special kind of dataset - a dataset containing subsets created after grouping as a column.- Most operations recognize regular dataset and grouped dataset and process data accordingly.
- One function form to enable thread-first on dataset.
Important! This library is not the replacement of tech.ml.dataset
nor a separate library. It should be considered as a addition on the top of tech.ml.dataset
.
If you want to know more about tech.ml.dataset
and dtype-next
please refer their documentation:
Join the discussion on Zulip
Please refer detailed documentation with examples.
The old documentation (till the end of 2023) is here.
(require '[tablecloth.api :as tc])
(-> "https://github.com/raw/techascent/tech.ml.dataset/master/test/data/stocks.csv"
(tc/dataset {:key-fn keyword})
(tc/group-by (fn [row]
{:symbol (:symbol row)
:year (tech.v3.datatype.datetime/long-temporal-field :years (:date row))}))
(tc/aggregate #(tech.v3.datatype.functional/mean (% :price)))
(tc/order-by [:symbol :year])
(tc/head 10))
_unnamed [10 3]:
:symbol | :year | summary |
---|---|---|
AAPL | 2000 | 21.74833333 |
AAPL | 2001 | 10.17583333 |
AAPL | 2002 | 9.40833333 |
AAPL | 2003 | 9.34750000 |
AAPL | 2004 | 18.72333333 |
AAPL | 2005 | 48.17166667 |
AAPL | 2006 | 72.04333333 |
AAPL | 2007 | 133.35333333 |
AAPL | 2008 | 138.48083333 |
AAPL | 2009 | 150.39333333 |
Tablecloth
is open for contribution. The best way to start is discussion on Zulip.
Documentation is written in the Kindly convention and is rendered using Clay composed with Quarto.
The old documentation was written in RMarkdown and is kept under docs/old/.
Documentation contains around 600 code snippets which are run during build. There are three relevant source files:
- README-source.md for README.md
- notebooks/index.clj for the detailed documentation
- clay.edn for some styling options of the docs
(notebooks/index.clj
was generated by dev/conversion.clj from the earlier Rmarkdown-based index.Rmd
with asome additional manual editing. Starting at 2024, it will diverge from that source, that will no longer be maintained.)
To generate README.md
, run the generate!
function at the dev/readme_generation.clj script.
To generate the detailed documentation, call the following. You will need the Quarto CLI installed in your system.
Currently (April 2024), we use Quarto's v1.5.10 pre-release (specifically this version, not the later ones) due to some Quarto bugs.
(require '[scicloj.clay.v2.api :as clay])
(clay/make! {:format [:quarto :html]
:source-path "notebooks/index.clj"})
To build this project fully we need to perform some code generation operations. These are listed below:
-
Build the
tablecloth.api.operators
namespaceThe
tablecloth.api.operators
namespace is generated bytablecloth.api.lift_operators
. To build that namespace, you need to load thetablecloth.api.lift_operators
namespace, and then execute the code surrounded by a comment at the bottom of the file. -
Build the
tablecloth.api
(aka the Dataset API)The
tablecloth.api
namespace is generated out ofapi-template
. To build that namespace you need to load thetablecloth.api.api-template
namespace, and then evaluate the code contained in the comment section at the bottom of the file. This will re-generate thetablecloth.api
namespace. -
Build the
tablecloth.column.api.operators
namespaceThe
tablecloth.column.api.operators
namespace is generated bytablecloth.column.api.lift_operators
. To build that namespace, you need to load thetablecloth.api.lift_operators
namespace, and then execute the code surrounded by a comment at the bottom of the file. -
Build the
tablecloth.column.api
(aka the Column API)The
tablecloth.column.api
namespace is generated out ofapi-template
. To build that namespace you need to load thetablecloth.column.api.api-template
namespace, and then evaluate the code contained in the comment section at the bottom of the file. This will re-generate thetablecloth.column.api
namespace.
- Before commiting changes please perform tests. I ususally do:
lein do clean, check, test
and build documentation as described above (which also tests whole library). - Keep API as simple as possible:
- first argument should be a dataset
- if parametrizations is complex, last argument should accept a map with not obligatory function arguments
- avoid variadic associative destructuring for function arguments
- usually function should working on grouped dataset as well, accept
parallel?
argument then (if applied).
- Follow
potemkin
pattern and import functions to the API namespace usingtech.v3.datatype.export-symbols/export-symbols
function - Functions which are composed out of API function to cover specific case(s) should go to
tablecloth.utils
namespace. - Always update
README-source.md
,CHANGELOG.md
,notebooks/index.clj
, tests and function docs are highly welcomed. - Always discuss changes and PRs first
Tests are written and run using midje. To run a test, evaluate a midje form. If it passes, it will return true
, if it fails details will be printed to the REPL.
- elaborate on tests
- tutorials
Copyright (c) 2020 Scicloj
The MIT Licence