Skip to content
LeetLLM
EasyEditing SystemsPython 3

File Patch Applier

Apply sorted or unsorted text edits atomically while rejecting invalid and overlapping ranges.

30m3 sample tests6 hidden tests

Implement apply_edits(text, edits), a safe text patch helper for non-overlapping edits.

Requirements

  • Each edit is (start, end, replacement) using half-open indexes from the selected language's string type. Python uses code-point indexes; Java String uses UTF-16 code-unit indexes.
  • before changing the text.
  • Raise ValueError for negative indexes, end < start, indexes past the text length, and overlapping edits.
  • For overlapping edits, the ValueError message must include the substring "overlap" (case-sensitive).
  • Adjacent edits are allowed because half-open ranges touching at an endpoint remain non-overlapping.
  • Multiple empty inserts at the same index are allowed. Preserve their input order so earlier inserts appear left-to-right in the result.
  • Edits may arrive unsorted.
  • Every offset addresses the original text, not the result of an earlier edit.
  • An empty insert strictly inside a nonempty edit's range is a conflict and must raise an overlap error. Inserts at either endpoint are allowed. At a replacement's start, insert text appears before its replacement; at its end, it appears after.
  • Apply all edits atomically and return the new text.

Example

Replace the half-open interval [6, 11), which covers world. The same original-coordinate rule supports edits at both ends, ordered inserts and Unicode:

python
1assert apply_edits("hello world", [(6, 11, "Cursor")]) == "hello Cursor" 2assert apply_edits("abcdef", [(0, 1, "A"), (5, 6, "F")]) == "AbcdeF" 3assert apply_edits("x", [(0, 0, "A"), (0, 0, "B")]) == "ABx" 4assert apply_edits("a🙂c", [(1, 2, "b")]) == "abc" # Python counts the emoji once

Keep original coordinates stable

For "abcdef", edits [(1, 3, "LONG"), (4, 6, "!")] replace bc and ef. The result is "aLONGd!". If you apply the first edit and then interpret [4, 6) against the changed string, you'll replace the wrong characters. A piece builder reads unchanged slices from the original text and appends replacements once.

Boundary inserts have an explicit ordering rule. Applying [(1, 2, "R"), (1, 1, "L"), (2, 2, "E")] to "abc" returns "aLREc". Inserting at index 2 inside replacement [1, 3) is rejected instead. Although an empty mathematical interval contains no characters, this API treats an interior insertion as an ambiguous edit conflict.

Python's [1, 2) covers the whole emoji in "a🙂c"; Java needs [1, 3) for that emoji because it occupies two UTF-16 units. Neither indexing convention promises that a range follows user-visible grapheme boundaries, such as a letter plus a combining accent. Use the language-specific offsets in this exercise.

Constraints

  • Use the selected language's native string indexes: Python code points or Java String UTF-16 code units.
  • Don't mutate the input.
  • Don't silently skip invalid edits.

Editor