Language Data Models

Based on "The Memory Models That Underlie Programming Languages" and my comments there.

C and C++: typed object graph, COBOL records, homogeneous arrays -- hm this is the fullest model.
Go: like C and C++ but with GC
Java: typed object graph + homogeneous arrays (people sort of viewed this as a "wart", but it looks good in retrospect)
JavaScript: untyped object graph. ES5 or ES6 added homogeneous arrays!
Python: untyped object graph only! Pandas adds homogeneous arrays. (Values have types, but the composites don't have types)
Lua: same as Python, but no Pandas
Perl: same as Python
shell: POSIX shell has no model -- just strings. Bash has half an array type and an associative array type which are sort of like an untyped COBOL records system. (He mentions the hierarchical file system as a model, which is shell's answer for more structured data, but that proves it doesn't belong in the taxonomy)
OCaml: typed object graph. It might have arrays, but typed linked lists are more central (and same with SML)
Scientific Languages
- R: untyped object graph + homogeneous arrays, tables of heterogeneous columns
- Julia/Matlab: matrices (multidimensional homogeneous arrays) are the core.
- Wolfram/Mathematica: symbols?
Jai language: like C, but has array-of-structures and structure-of-arrays flexibility. Logical vs. Physical data model.

JSON
Protocol Buffers. NOTE: Added both maps and unions (could be mapped to sum types)
ASDL

Systems that Persist Data

SQL: relational model persisted
- concrete storage: could be row-oriented or column oriented
File system
Redis: dicts, lists, sets, counters (I think)