Skip to content
Quannt's Technical Blog
SpeakerDeckGithubTwitterPortfolioLinkedin

How to write good comments

software3 min read

Why write comments

  • "But my code is perfect", no.
  • Provide contexts, the whats, the whys, abstractions, and public interfaces for the code.

“comments should describe things that aren’t obvious from the code.” “Developers should be able to understand the abstraction provided by a module without reading any code other than its externally visible declarations”

Excerpt From: John Ousterhout. “A Philosophy of Software Design.”

Types of comments

  1. Lower-level comments that add precision to the code.
//  Position in this buffer of the first object that hasn't
//  been returned to the client.
uint32_t offset;
//  Holds statistics about line lengths of the form <length, count>
//  where length is the number of characters in a line (including
//  the newline), and count is the number of lines with
//  exactly that many characters. If there are no lines with
//  a particular length, then there is no entry for that length.
private TreeMap<Integer, Integer> numLinesWithLength;
  1. Higher-level comments that add high-level information to the code. Good comments answer these questions: what is it that the code trying to achieve here? What's the most important thing about this code?
// Try to append the current key hash onto an existing
// RPC to the desired server that hasn't been sent yet.”
int readActiveRpcId = RPC_ID_NOT_ASSIGNED;
for (int i = 0; i < NUM_READ_RPC; i++) {
      if (session == readRpc[i].session
                 && readRpc[i].status == LOADING
                 && readRpc[i].maxPos < assignPos
                 && readRpc[i].numHashes < MAX_PKHASHES_PERRPC) {
          readActiveRpcId = i;
          break;
      }
}
  1. Interface-level comments define the public contract of the code without telling the implementation details.

Bad example

/* This class implements the client side framework for index range
 * lookups. It manages a single LookupIndexKeys RPC and multiple
 * IndexedRead RPCs. Client side just includes "IndexLookup.h" in
 * its header to use IndexLookup class. Several parameters can be set
 * in the config below:
 * - The number of concurrent indexedRead RPCs
 * - The max number of PKHashes a indexedRead RPC can hold at a time
 * - The size of the active PKHashes
 *
 * To use IndexLookup, the client creates an object of this class by
 * providing all necessary information. After construction of
 * IndexLookup, client can call getNext() function to move to next
 * available object. If getNext() returns NULL, it means we reached
 * the last object. Client can use getKey, getKeyLength, getValue,
 * and getValueLength to get object data of current object.
 */

Good example

/*
 * This class is used by client applications to make range queries
 * using indexes. Each instance represents a single range query.
 *
 * To start a range query, a client creates an instance of this
 * class. The client can then call getNext() to retrieve the objects
 * in the desired range. For each object returned by getNext(), the
 * caller can invoke getKey(), getKeyLength(), getValue(), and
 * getValueLength() to get information about that object.
*
* \param offset
 *        Index within the buffer of the first byte to copy.
 * \param length
 *        Number of bytes to copy.
 * \param dest
 *        Where to copy the bytes: must have room for at least
 *        length bytes.
 *
 * \return
 *        The return value is the actual number of bytes copied,
 *        which may be less than length if the requested range of
 *        bytes extends past the end of the buffer. 0 is returned
 *        if there is no overlap between the requested range and
 *        the actual buffer.
 */

Excerpt From: John Ousterhout. “A Philosophy of Software Design.”

© 2024 by Quannt's Technical Blog. All rights reserved.