This page uses content from Wikipedia and is licensed under CC BY-SA.

Piece table

In computing, a piece table is a data structure typically used to represent a series of edits on a text document. An initial reference (or 'span') to the whole of the original file is created, with subsequent inserts and deletes being created as combinations of one, two, or three references to sections of either the original document or of the spans associated with earlier inserts.[1]

Typically the text of the original document is held in one immutable block, and the text of each subsequent insert is stored in new immutable blocks. Because even deleted text is still included in the piece table, this makes multi-level or unlimited undo easier to implement with a piece table than with alternative data structures such as a gap buffer.

This data structure is invented by J Strother Moore.[2]


For this description, we use buffer as the immutable block to held the contents.

A piece table[1] consists of three columns:

  • Which buffer
  • Start index in the buffer
  • Length in the buffer

In addition to the table, two buffers are used:

  • "Original buffer": A buffer to the original text document. This buffer is read-only.
  • "Add buffer": A buffer to a temporary file. This buffer is append-only.



Definition: Index(i): return the character at position i

To retrieve the i-th character, the appropriate entry in a piece table is read.


Given the following buffers and piece table:

Buffer Content
Original file ipsum sit amet
Add file Lorem deletedtext dolor
Piece table
Which Start Index Length
Add 0 6
Original 0 6
Add 18 5
Original 6 9

To access the i-th character, the appropriate entry in the piece table is looked up.

For instance, to get the value of Index(15), the 3rd entry of piece table is retrieved. This is because the 3rd entry describes the characters from index 12 to 17. (the first entry describes characters in index 0 to 5, the next one is 6 to 11). The piece table entry simply means to look up at the "add file" buffer, starting at index 18 in that buffer. The relative index in that entry is 3 (15 - 12), hence the value of Index(15) is "o".

For the buffers and piece table given above, the following text is currently being shown

Lorem ipsum dolor sit amet


Inserting characters to the text consists of:

  • Appending characters to the "add file" buffer, and
  • Updating the entry in piece table (breaking an entry into two or three)


Deletion involves only modifying the appropriate entry in piece table.

Time complexity

Although piece table itself is a data structure, it does not states how it should be implemented. The time complexity of operations depends on how the table is being implemented. One possible implementation of piece table is in a splay tree, because of quick access of recently-accessed elements.


Several text editors use an in-RAM piece table internally, including Bravo,[1] Abiword[3][4][5], Atom[6] and Visual Studio Code.[7]

The "fast save" feature in some versions of Microsoft Word uses a piece table for the on-disk file format.[2]

The on-disk representation of text files in the Oberon System[further explanation needed] uses a piece chain technique that allows pieces of one document to point to text stored in some other document, similar to transclusion. [8]


  1. ^ a b c Crowley, Charles (1998-06-10). "Data Structures for Text Sequences - 6.4 The piece table method" (PDF). Archived from the original (PDF) on 2018-02-23.
  2. ^ a b David Lu. "What's been wrought using the Piece Table?". (discussion)
  3. ^ "AbiWord Development: Piece Table Background".
  4. ^ James Brown. "Piece Chains: Design & Implementation of a Win32 Text Editor".
  5. ^ Joaquin Cuenca Abela. "Improving the AbiWord's Piece Table".
  6. ^ "Atom's new concurrency-friendly buffer implementation"
  7. ^ "VS Code 1.21 Release Notes
  8. ^ Niklaus Wirth, Jürg Gutknecht. "Project Oberon: The Design of an Operating System and Compiler". 2005. p. 90.