SQLite 终端常用命令

2026-05-18T00:00:00+00:00
SQLite 自带的 sqlite3</code> 命令行工具很适合临时查数据、排查问题、导入导出小文件。下面整理一份常用命令，日常够用了。</p>
启动和打开数据库#</a>
</h2>
# 打开已有数据库；文件不存在时会自动创建
</span>sqlite3</span> app.db
</span>
</span># 只读方式打开，适合线上排查
</span>sqlite3 -readonly</span> app.db
</span>
</span># 直接执行一条 SQL 后退出
</span>sqlite3</span> app.db "</span>select count(*) from users;</span>"
</span>
</span># 打开内存数据库
</span>sqlite3</span> :memory:
</span></code></pre>
进入交互终端后，提示符一般是：</p>
sqlite>
</span></code></pre>
SQLite 终端里有两类命令：</p>

SQL 语句：例如 select * from users;</code>，末尾需要分号。</li>
点命令：例如 .tables</code>、.schema</code>，以 .</code> 开头，不需要分号。</li>
</ul>
查看帮助和退出#</a>
</h2>
.help       </span>-- 查看所有点命令
</span>.quit       </span>-- 退出
</span>.exit       </span>-- 退出
</span></code></pre>
如果 SQL 写了一半进入了多行模式，可以用 Ctrl+C</code> 取消当前输入。</p>
查看数据库结构#</a>
</h2>
.databases              </span>-- 查看当前连接的数据库文件
</span>.tables                 </span>-- 查看所有表
</span>.tables user%           </span>-- 按模式过滤表名
</span>.schema                 </span>-- 查看完整建表语句
</span>.schema users           </span>-- 查看某张表的建表语句
</span>.indexes                </span>-- 查看所有索引
</span>.indexes users          </span>-- 查看某张表的索引
</span></code></pre>
也可以直接查系统表：</p>
select</span> name, type
</span>from</span> sqlite_master
</span>where</span> type in ('</span>table</span>', '</span>index</span>', '</span>view</span>')
</span>order by</span> type, name;
</span></code></pre>
查看表字段：</p>
pragma table_info(users);
</span></code></pre>
查看外键：</p>
pragma foreign_key_list(users);
</span></code></pre>
输出格式设置#</a>
</h2>
默认输出不太适合阅读，建议先打开表头和列模式：</p>
.headers on
</span>.mode column
</span>select</span> id, name, created_at </span>from</span> users </span>limit </span>10</span>;
</span></code></pre>
常用输出模式：</p>
.mode column       </span>-- 表格列对齐，适合人看
</span>.mode </span>box          </span>-- 边框表格，新版本 sqlite3 更好看
</span>.mode list         </span>-- 默认模式
</span>.mode csv          </span>-- CSV
</span>.mode json         </span>-- JSON
</span>.mode </span>line         </span>-- 每列一行，适合字段很多的记录
</span></code></pre>
设置空值显示：</p>
.nullvalue </span>NULL
</span></code></pre>
打开 SQL 执行时间：</p>
.timer on
</span></code></pre>
常用查询#</a>
</h2>
查看前几行：</p>
select </span>* </span>from</span> users </span>limit </span>10</span>;
</span></code></pre>
按时间倒序：</p>
select </span>*
</span>from</span> users
</span>order by</span> created_at </span>desc
</span>limit </span>20</span>;
</span></code></pre>
统计行数：</p>
select </span>count</span>(</span>*</span>) </span>from</span> users;
</span></code></pre>
模糊搜索：</p>
select </span>*
</span>from</span> users
</span>where</span> name like '</span>%tom%</span>';
</span></code></pre>
查看重复数据：</p>
select</span> email, </span>count</span>(</span>*</span>) as n
</span>from</span> users
</span>group by</span> email
</span>having n > </span>1</span>;
</span></code></pre>
删除数据#</a>
</h2>
删除前先查一遍，确认条件命中了正确的数据：</p>
select </span>*
</span>from</span> users
</span>where</span> id = </span>10</span>;
</span></code></pre>
删除一条记录：</p>
delete from</span> users
</span>where</span> id = </span>10</span>;
</span></code></pre>
按条件删除多条记录：</p>
delete from</span> users
</span>where</span> status = '</span>disabled</span>';
</span></code></pre>
删除前统计影响行数：</p>
select </span>count</span>(</span>*</span>)
</span>from</span> users
</span>where</span> status = '</span>disabled</span>';
</span></code></pre>
更稳妥的做法是放进事务里：</p>
begin</span>;
</span>
</span>delete from</span> users
</span>where</span> status = '</span>disabled</span>';
</span>
</span>-- 确认删除后的结果
</span>select </span>count</span>(</span>*</span>) </span>from</span> users;
</span>
</span>commit</span>;
</span></code></pre>
如果发现删错了，在 commit;</code> 之前可以回滚：</p>
rollback</span>;
</span></code></pre>
清空整张表：</p>
delete from</span> users;
</span></code></pre>
如果表有自增主键，并且想同时重置自增序列：</p>
delete from</span> users;
</span>delete from</span> sqlite_sequence </span>where</span> name = '</span>users</span>';
</span></code></pre>
SQLite 没有 MySQL 那种 truncate table users;</code>，清空表一般就是用 delete from users;</code>。</p>
删除表结构和数据：</p>
drop table</span> users;
</span></code></pre>
drop table</code> 会把整张表删掉，表结构也没了，执行前最好先 .schema users</code> 或 .dump users</code> 留一份。</p>
导入和导出#</a>
</h2>
导出查询结果到 CSV#</a>
</h3>
.headers on
</span>.mode csv
</span>.once </span>users</span>.</span>csv
</span>select</span> id, name, email </span>from</span> users;
</span></code></pre>
.once</code> 只影响下一条 SQL。也可以用 .output</code> 持续输出到文件：</p>
.output </span>users</span>.</span>csv
</span>select </span>* </span>from</span> users;
</span>.output stdout
</span></code></pre>
导入 CSV 到表#</a>
</h3>
先确认表已经存在：</p>
create table </span>users_import</span> (
</span>    id </span>integer</span>,
</span>    name </span>text</span>,
</span>    email </span>text
</span>);
</span></code></pre>
再导入：</p>
.mode csv
</span>.import </span>users</span>.</span>csv</span> users_import
</span></code></pre>
如果 CSV 第一行是表头，新版 sqlite3 可以这样跳过：</p>
.import </span>--skip 1 users.csv users_import
</span></code></pre>
导出整个数据库#</a>
</h3>
sqlite3</span> app.db "</span>.dump</span>" > app.sql
</span></code></pre>
只导出某张表：</p>
sqlite3</span> app.db "</span>.dump users</span>" > users.sql
</span></code></pre>
从 SQL 文件恢复：</p>
sqlite3</span> new.db < app.sql
</span></code></pre>
备份和复制数据库#</a>
</h2>
在 sqlite3 交互终端中备份：</p>
.backup </span>backup</span>.</span>db
</span></code></pre>
把当前数据库保存为新文件：</p>
.save </span>copy</span>.</span>db
</span></code></pre>
如果数据库正在被应用使用，优先用 .backup</code>，不要直接复制 app.db</code> 文件。</p>
执行 SQL 文件#</a>
</h2>
sqlite3</span> app.db < migration.sql
</span></code></pre>
或在交互终端里执行：</p>
.read </span>migration</span>.</span>sql
</span></code></pre>
事务操作#</a>
</h2>
手动开启事务：</p>
begin</span>;
</span>
</span>update</span> users
</span>set</span> status = '</span>disabled</span>'
</span>where</span> last_login_at < '</span>2025-01-01</span>';
</span>
</span>commit</span>;
</span></code></pre>
发现不对就回滚：</p>
rollback</span>;
</span></code></pre>
线上手动改数据时，建议先 begin;</code>，查清影响范围后再 commit;</code>。</p>
查看执行计划#</a>
</h2>
explain query plan
</span>select </span>*
</span>from</span> users
</span>where</span> email = '</span>a@example.com</span>';
</span></code></pre>
如果结果里看到全表扫描，并且这条查询很频繁，可以考虑加索引：</p>
create index </span>idx_users_email on </span>users</span>(email);
</span></code></pre>
常用 PRAGMA#</a>
</h2>
查看 SQLite 版本：</p>
select</span> sqlite_version();
</span></code></pre>
查看数据库页大小和大小信息：</p>
pragma page_size;
</span>pragma page_count;
</span>pragma freelist_count;
</span></code></pre>
开启外键约束：</p>
pragma foreign_keys = on;
</span></code></pre>
查看外键是否开启：</p>
pragma foreign_keys;
</span></code></pre>
检查数据库完整性：</p>
pragma integrity_check;
</span></code></pre>
清理空间：</p>
vacuum;
</span></code></pre>
更新统计信息，帮助查询优化器：</p>
analyze;
</span></code></pre>
Attach 多个数据库#</a>
</h2>
有时需要在两个 SQLite 数据库之间复制数据：</p>
attach database '</span>old.db</span>' as old;
</span>
</span>insert into</span> users (id, name, email)
</span>select</span> id, name, email
</span>from </span>old</span>.</span>users</span>;
</span>
</span>detach database old;
</span></code></pre>
查看当前连接：</p>
.databases
</span></code></pre>
排查锁问题#</a>
</h2>
如果遇到：</p>
database is locked
</span></code></pre>
可以先设置等待时间：</p>
.timeout </span>5000
</span></code></pre>
表示最多等待 5000 毫秒。</p>
也可以检查当前 journal 模式：</p>
pragma journal_mode;
</span></code></pre>
开发环境可以考虑启用 WAL：</p>
pragma journal_mode = wal;
</span></code></pre>
WAL 通常能改善读写并发，但是否适合线上环境要结合部署方式判断，尤其是网络文件系统上要谨慎。</p>
一个常用启动模板#</a>
</h2>
平时查库可以这样开：</p>
sqlite3 -readonly</span> app.db
</span></code></pre>
进入后先执行：</p>
.headers on
</span>.mode </span>box
</span>.timer on
</span>.nullvalue </span>NULL
</span>.tables
</span></code></pre>
如果 sqlite3 版本不支持 .mode box</code>，就换成：</p>
.mode column
</span></code></pre>
小结#</a>
</h2>
最常用的一组命令：</p>
.help
</span>.databases
</span>.tables
</span>.schema users
</span>.headers on
</span>.mode </span>box
</span>.timer on
</span>.once </span>result</span>.</span>csv
</span>.import </span>data</span>.</span>csv</span> table_name
</span>.backup </span>backup</span>.</span>db
</span>.read </span>file</span>.</span>sql
</span>.quit
</span></code></pre>
记住点命令不加分号、SQL 语句要加分号，基本就能顺畅使用 sqlite3 终端了。</p>
orhan.cn - sqlite

SQLite 终端常用命令
