| # Fixed 2-letter aliases for common Python attribute names. | |
| # | |
| # Any `Attribute.attr` node whose name appears here is emitted as the corresponding 2-letter | |
| # alias. Attrs not in this table fall back to being emitted as a raw single-letter stream, | |
| # which often pushes the sample past the 128-token cap and gets it filtered out. | |
| # | |
| # Invariants enforced by the Vocab loader: | |
| # 1. No alias equals a 2-letter Python keyword (`as`, `if`, `in`, `is`, `or`). | |
| # 2. No alias appears in both this file and `builtin_aliases.yaml`. | |
| # | |
| # Aliases are picked as distinctive mnemonics; collisions resolved by picking distinct | |
| # second letters within a functional group. | |
| # --- list / sequence methods --- | |
| append: aa | |
| extend: xt | |
| insert: ti # "in" is a keyword, use ti (ins-erT + I) | |
| remove: mv | |
| clear: cw | |
| count: ct | |
| # --- dict / set methods --- | |
| update: ud | |
| setdefault: sf | |
| pop: pp | |
| popitem: pi | |
| items: jm # "it" would collide with iter's alias; use j as mnemonic for pl-j-urals | |
| keys: kk | |
| get: gu | |
| values: vu | |
| # --- sorting / reversing (distinct from builtins sorted/reversed) --- | |
| sort: so | |
| reverse: vv # "rv" is taken by reversed() | |
| # --- string methods --- | |
| find: fi | |
| split: pl | |
| join: jn | |
| strip: tq | |
| lower: lw | |
| upper: ur | |
| replace: rl | |
| startswith: sw | |
| endswith: ew | |
| encode: nc | |
| decode: dk # "de" feels too generic | |
| format: ft | |
| # --- file / io methods --- | |
| read: rd | |
| write: wt | |
| readline: rq | |
| readlines: rw | |
| close: cg | |
| seek: sq | |
| tell: tu | |
| flush: fu | |
| # --- object attributes --- | |
| name: nm | |
| value: vl | |
| data: da | |
| size: sz | |
| length: lh | |
| shape: sk | |
| dtype: dy | |
| # --- common numeric attrs (ndarray / tensor) --- | |
| mean: mj | |
| std: sj | |
| var: va | |