From fd424eceb00e4aa535f2e312c4f1e9957a2cac84 Mon Sep 17 00:00:00 2001 From: alexey-milovidov Date: Wed, 14 Apr 2021 23:02:28 +0300 Subject: [PATCH] Update code-review.md --- website/blog/en/2021/code-review.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/website/blog/en/2021/code-review.md b/website/blog/en/2021/code-review.md index 6336d132317..dcde371629b 100644 --- a/website/blog/en/2021/code-review.md +++ b/website/blog/en/2021/code-review.md @@ -62,7 +62,7 @@ It is common to hear objections to the idea of commenting the code, so let's dis ### Self-documenting Code -You can often see a perplexing idea that the source code can somehow be "self-documenting", or that the comments are a "code smell", and their presence indicates that the code is badly written. I have trouble imagining how this belief can be compatible with any experience in maintaining sufficiently complex and large software, over the years, in collaboration with others. The code and the comments describe different parts of the solution. The code describes the data structures and their transformations, but it cannot convey meaning. The names in the code serve as pointers that map the data and its transformations to the domain concepts, but they are schematic and lack nuance. It is not so difficult to write code that makes it easy to understand what's going on in terms of data manipulation. What it takes is mostly moderation, that is, stopping yourself from being too clever. For most code, it is easy to see what it does, but why? Why this way and not that way? Why is it correct? Why this fast path here helps? Why did you choose this data layout? How is this invariant guaranteed? And so on. This might be not so evident for a developer who is working alone on a short-lived project, because they have all the necessary context in their head. But when they have to work with other people (or even with themselves from past and future), or in an unfamiliar area, the importance of non-code, higher-level context becomes painfully clear. The idea that we should, or even can, somehow encode comments such as [this one](https://github.com/ClickHouse/ClickHouse/blob/495c6e03aa9437dac3cd7a44ab3923390bef9982/src/Storages/MergeTree/KeyCondition.cpp#L1312-L1347) into names or control flow is just absurd. +You can often see a perplexing idea that the source code can somehow be "self-documenting", or that the comments are a "code smell", and their presence indicates that the code is badly written. I have trouble imagining how this belief can be compatible with any experience in maintaining sufficiently complex and large software, over the years, in collaboration with others. The code and the comments describe different parts of the solution. The code describes the data structures and their transformations, but it cannot convey meaning. The names in the code serve as pointers that map the data and its transformations to the domain concepts, but they are schematic and lack nuance. It is not so difficult to write code that makes it easy to understand what's going on in terms of data manipulation. What it takes is mostly moderation, that is, stopping yourself from being too clever. For most code, it is easy to see what it does, but why? Why this way and not that way? Why is it correct? Why this fast path here helps? Why did you choose this data layout? How is this invariant guaranteed? And so on. This might be not so evident for a developer who is working alone on a short-lived project, because they have all the necessary context in their head. But when they have to work with other people (or even with themselves from past and future), or in an unfamiliar area, the importance of non-code, higher-level context becomes painfully clear. The idea that we should, or even can, somehow encode comments such as [this one](https://github.com/ClickHouse/ClickHouse/blob/26d5db32ae5c9f54b8825e2eca1f077a3b17c84a/src/Storages/MergeTree/KeyCondition.cpp#L1312-L1347) into names or control flow is just absurd. ### Obsolete Comments