Summary

Vera's string built-ins cover the most common operations but are missing several utilities that come up frequently in text processing and formatting. Two distinct gaps:

1. Character-level ops (string_reverse, string_pad_*, string_chars, string_trim_*) — close the per-character recursion pattern.
2. Structural splits (string_lines, string_words) — string_split only takes a single delimiter character, so splitting on line endings (handling \r\n) or whitespace runs requires regex today.

Proposed API

Transformation

• string_reverse(@String) -> @String — reverse a string (Unicode-aware, by grapheme cluster)
• string_pad_start(@String, @Nat, @String) -> @String — pad to target length with fill string (left-pad)
• string_pad_end(@String, @Nat, @String) -> @String — pad to target length with fill string (right-pad)

Decomposition

• string_chars(@String) -> @Array<String> — split into individual characters (grapheme clusters) as an array of single-character strings. Bridge primitive: once this exists, string operations become array operations (e.g. string_reverse = string_from_chars(array_reverse(string_chars(s))), character predicates become array_any(string_chars(s), is_digit)).

Structural splits

• string_lines(@String) -> @Array<String> — split on line terminators (\n, \r\n, \r). Mirrors Python's str.splitlines(), Rust's str::lines(), Java's String.lines(). Current workaround requires string_split("\n") and manual \r stripping — LLMs get this wrong on Windows-style line endings.
• string_words(@String) -> @Array<String> — split on whitespace runs (spaces, tabs, newlines), discarding empty segments. Mirrors Python's str.split() with no argument. Currently impossible with string_split (single-delimiter only); requires regex.

Trimming variants

• string_trim_start(@String) -> @String — remove leading whitespace only
• string_trim_end(@String) -> @String — remove trailing whitespace only

Note: string_strip already exists for trimming both sides.

Implementation

• environment.py: Register as pure functions
• codegen/api.py: Host imports delegating to Python string methods (.ljust(), .rjust(), .lstrip(), .rstrip(), list() for chars, .splitlines() for lines, .split() with no arg for words)
• Browser runtime: padStart(), padEnd(), trimStart(), trimEnd(), spread operator for chars, regex-based split for lines/words
• Verification:
  • string_reverse preserves length
  • string_pad_start / string_pad_end: result length >= input length
  • string_trim_*: result length <= input length
  • string_chars: result length = string_length of input
  • string_lines / string_words: concatenating results (with separator) recovers a prefix of the input

Priority

Medium. Less urgent than array utilities, sleep, and random. Within this issue, priority order:

1. string_chars — bridge primitive; unlocks array-combinator approach to string processing
2. string_lines / string_words — real capability gaps; string_split can't currently do these correctly
3. string_pad_start / string_pad_end — common formatting need
4. string_reverse, string_trim_start, string_trim_end — convenience, once string_chars exists several of these become one-liners