数据库中有多个状态字段的存储决策（bit vs boolean vs integer）

在可以拆分定义时，优先将业务拆分字段定义，可以boolean时优先使用boolean。

否则考虑使用integer。或使用bit（但是会带来可读性成本）。

摘自StackOverflow，以下为原文。

Q：

I am working with a table in a PostgreSQL database that has several boolean columns that determine some state (e.g. published, visible, etc.). I want to make a single status column that will store all these values as well as possible new ones in a form of a bitmask. Is there any difference between integer and bit(n) in this case?

This is going to be a rather big table, because it stores objects that users create via a web-interface. So I think I will have to use (partial) indexes for this column.

A：

If you only have a few variables I would consider keeping separate boolean columns.

• Indexing is easy. In particular also indexes on expressions and partial indexes.
• Conditions for queries are easy to write and read and meaningful.
• A boolean column occupies 1 byte (no alignment padding). For only a few variables this occupies the least space.
• Unlike other options boolean columns allow NULL values for individual bits if you should need that. You can always define columns NOT NULL if you don't.

If you have more than a hand full variables but no more than 32, an integer column may serve best. (Or a bigint for up to 64 variables.)

• Occupies 4 bytes on disk (may require alignment padding, depending on preceding columns).
• Very fast indexing for exact matches ( = operator).
• Handling individual values may be slower / less convenient than with varbit or boolean.

With even more variables, or if you want to manipulate the values a lot, or if you don't have huge tables or disk space / RAM is not an issue, or if you are not sure what to pick, I would consider bit(n) or bit varying(n) (short: varbit(n).

• Occupies at least 5 bytes (or 8 for very long strings) plus 1 byte for each group of 8 bits (rounded up).
• You can use bit string functions and operators directly, and some standard SQL functions as well.

For just 3 bits of information, individual boolean columns get by with 3 bytes, an integer needs 4 bytes (maybe additional alignment padding) and a bit string 6 bytes (5 + 1).

For 32 bits of information, an integer still needs 4 bytes (+ padding), a bit string occupies 9 bytes for the same (5 + 4) and boolean columns occupy 32 bytes.

To optimize disk space further you need to understand the storage mechanisms of PostgreSQL, especially data alignment. More in this related answer.

This answer on how to transform the types boolean, bit(n) and integer may be of help, too.