From bacac696d00fd6bd5c0968c25e6146fb8b760336 Mon Sep 17 00:00:00 2001 From: tangenta Date: Tue, 22 Nov 2022 18:19:50 +0800 Subject: [PATCH] docs: fix typo and fix dead link --- docs/design/2020-01-24-collations.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/design/2020-01-24-collations.md b/docs/design/2020-01-24-collations.md index a222035745b83..af7c9e30b2f26 100644 --- a/docs/design/2020-01-24-collations.md +++ b/docs/design/2020-01-24-collations.md @@ -105,10 +105,10 @@ The interface is quite similar to the Go [collate package](https://godoc.org/gol ### Row Format -The encoding layout of TiDB has been described in our [previous article](https://pingcap.com/blog/2017-07-11-tidbinternal2/#map). The row format should be changed to make it memory comparable, this is important to the index lookup. Basic principle is that all keys encoded for strings should use the `sortKeys` result from `Key()`/`KeyFromString()` function. However, most of the `sortKeys` calculations are not reversible. +The encoding layout of TiDB has been described in our [previous article](https://docs.pingcap.com/tidb/stable/tidb-computing). The row format should be changed to make it memory comparable, this is important to the index lookup. Basic principle is that all keys encoded for strings should use the `sortKeys` result from `Key()`/`KeyFromString()` function. However, most of the `sortKeys` calculations are not reversible. * For table data, encodings stay unchanged. All strings are compared after decoding with the `Compare()` function. - * For table indices, we replace current `ColumnValue` with `sortKey` and encode the `ColumnValue` to the value,: + * For table indices, we replace current `ColumnValue` with `sortKey` and encode the `ColumnValue` to the value: - For unique indices: ``` Key: tablePrefix{tableID}_indexPrefixSep{indexID}_sortKey