Records with Rank Polymorphism
summary
Remora implements rank-polymorphism and the team embeds a dataframe library to show how columnar tables map to vectors of records, using Racket, Standard ML and Python as comparisons. The prototype builds dataframes row-wise and column-wise, supports projection, update and row filtering by composing lenses (view/set/over), and handles ragged results with boxed arrays. Racket structs prove awkward; the synthesized record constructor plus lens design is practical and interoperates cleanly with lifted array ops.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:34The Remora Project
- 2:03Two kinds of Aggregate Dota
- 4:53Data Model
- 7:36Function Application
- 19:55Lenses
- 20:51Lens Operations
- 21:50Record Creation
- 22:50Column Extraction
related talks
25mAn Array-Oriented Language via the Design Recipe
32mPandas, Polars and the DataFrame Consortium.
51mProgramming in the Tidyverse
40mEnter the Matrix
43mInteroperability between R Language & other Languages
33mPortable, type-fancy multidimensional arrays
26mDoing data science with Clojure: the ugly, the sad, the joyful
31mBeyond Pandas: lightning fast in-memory dataframes with Polars