简介#

如果说要给Linux文本三剑客(grep、sed、awk)添加一员的话，我觉得应该是jq命令，因为jq命令是用来处理json数据的工具，而现如今json几乎无所不在！

网上的jq命令分享文章也不少，但大多介绍得非常浅，jq的强大之处完全没有介绍出来，所以就有了这篇文章，安利一下jq这个命令。

基本用法#

格式化#

# jq默认的格式化输出
$ echo -n '{"id":1, "name":"zhangsan", "score":[75, 85, 90]}'|jq .
{
  "id": 1,
  "name": "zhangsan",
  "score": [
    75,
    85,
    90
  ]
}

# -c选项则是压缩到1行输出
$ jq -c . <<eof
{
  "id": 1,
  "name": "zhangsan",
  "score": [
    75,
    85,
    90
  ]
}
eof
{"id":1,"name":"zhangsan","score":[75,85,90]}

属性提取#

# 获取id字段
$ echo -n '{"id":1, "name":"zhangsan", "score":[75, 85, 90]}'|jq '.id'
1
# 获取name字段
$ echo -n '{"id":1, "name":"zhangsan", "score":[75, 85, 90]}'|jq '.name'
"zhangsan"

# 获取name字段，-r 解开字符串引号
$ echo -n '{"id":1, "name":"zhangsan", "score":[75, 85, 90]}'|jq -r '.name'
zhangsan

# 多层属性值获取
$ echo -n '{"id":1, "name":"zhangsan", "attr":{"height":1.78,"weight":"60kg"}}'|jq '.attr.height'
1.78

# 获取数组中的值
$ echo -n '{"id":1, "name":"zhangsan", "score":[75, 85, 90]}'|jq -r '.score[0]'
75

$ echo -n '[75, 85, 90]'|jq -r '.[0]'
75

# 数组截取
$ echo -n '[75, 85, 90]'|jq -r '.[1:3]'
[
  85,
  90
]

# []展开数组
$ echo -n '[75, 85, 90]'|jq '.[]'
75
85
90

# ..展开所有结构
$ echo -n '{"id":1, "name":"zhangsan", "score":[75, 85, 90]}'|jq -c '..'
{"id":1,"name":"zhangsan","score":[75,85,90]}
1
"zhangsan"
[75,85,90]
75
85
90

# 从非对象类型中提取字段，会报错
$ echo -n '{"id":1, "name":"zhangsan", "attr":{"height":1.78,"weight":"60kg"}}'|jq '.name.alias'
jq: error (at <stdin>:0): Cannot index string with string "alias"

# 使用?号可以避免这种报错
$ echo -n '{"id":1, "name":"zhangsan", "attr":{"height":1.78,"weight":"60kg"}}'|jq '.name.alias?'

# //符号用于，当前面的表达式取不到值时，执行后面的表达式
$ echo -n '{"id":1, "name":"zhangsan", "attr":{"height":1.78,"weight":"60kg"}}'|jq '.alias//.name'
"zhangsan"

管道、逗号与括号#

# 管道可以将值从前一个命令传送到后一个命令
$ echo -n '{"id":1, "name":"zhangsan", "attr":{"height":1.78,"weight":"60kg"}}'|jq '.attr|.height'
1.78

# jq中做一些基础运算也是可以的
$ echo -n '{"id":1, "name":"zhangsan", "attr":{"height":1.78,"weight":"60kg"}}'|jq '.attr|.height*100|tostring + "cm"'
"178cm"

# 逗号使得可以执行多个jq表达式，使得一个输入可计算出多个输出结果
$ echo 1 | jq '., ., .'
1
1
1

# 括号用于提升表达式的优先级，如下：逗号优先级低于算术运算
$ $ echo '1'|jq '.+1, .*2'
2
2

$ echo '1'|jq '(.+1, .)*2'
4
2

# 管道优先级低于逗号
$ echo '1'|jq '., .|tostring'
"1"
"1"

$ echo '1'|jq '., (.|tostring)'
1
"1"

理解jq执行过程#

表面上jq是用来处理json数据的，但实际上jq能处理的是任何json基础元素所形成的流，如integer、string、bool、null、object、array等，jq执行过程大致如下：

1. jq从流中获取一个json元素
2. jq执行表达式，表达式生成新的json元素
3. jq将新的json元素打印输出

可以看看这些示例，如下：

# 这里jq实际上将1 2 3 4当作4个integer元素，每找到一个元素就执行+1操作
# jq实际上是流式处理的，1 2 3 4可以看成流中的4个元素
$ echo '1 2 3 4'|jq '. + 1'
2
3
4
5

# 流中的元素不需要是同种类型，只要是完整的json元素即可
$ jq '"<" + tostring + ">"' <<eof
1
"zhangsan"
true
{"id":1}
[75, 80, 85]
eof

"<1>"
"<zhangsan>"
"<true>"
"<{\"id\":1}>"
"<[75,80,85]>"

# -R选项可用于将读取到的json元素，都当作字符串对待
$ seq 4|jq -R '.'
"1"
"2"
"3"
"4"

# -s选项将从流中读取到的所有json元素，变成一个json数组元素  
# 这里理解为jq从流中只取到了1个json元素，这个json元素的类型是数组
$ seq 4|jq -s .
[
  1,
  2,
  3,
  4
]

基础运算#

jq支持+ - * / %运算，对于+号，如果是字符串类型，则是做字符串拼接，如下：

# 做加减乘除运算
$ echo 1|jq '.+1, .-1, .*2, ./2, .%2'
2
0
2
0.5
1

# 赋值运算
$ echo -n '{"id":1,"name":"zhangsan","age":"17","score":"75"}'|jq '.id=2' -c
{"id":2,"name":"zhangsan","age":"17","score":"75"}

数据构造#

jq可以很方便的将其它数据，转化为json对象或数组，如下：

# 使用[]构造数组元素，-n告诉jq没有输入数据，直接执行表达式并生成输出数据
$ jq -n '[1,2,3,4]' -c
[1,2,3,4]

$ cat data.txt
id  name      age  score
1   zhangsan  17   75
2   lisi      16   80
3   wangwu    18   85
4   zhaoliu   18   90

# 每行分割成数组，[]构造新的数组输出
$ tail -n+2 data.txt|jq -R '[splits("\\s+")]' -c
["1","zhangsan","17","75"]
["2","lisi","16","80"]
["3","wangwu","18","85"]
["4","zhaoliu","18","90"]

$ jq -n '{id:1, name:"zhangsan"}' -c
{"id":1,"name":"zhangsan"}

# 每行转换为对象，{}构造新的对象格式输出
$ tail -n+2 data.txt|jq -R '[splits("\\s+")] | {id:.[0]|tonumber, name:.[1], age:.[2], score:.[3]}' -c
{"id":1,"name":"zhangsan","age":"17","score":"75"}
{"id":2,"name":"lisi","age":"16","score":"80"}
{"id":3,"name":"wangwu","age":"18","score":"85"}
{"id":4,"name":"zhaoliu","age":"18","score":"90"}

# \()字符串占位变量替换
$ cat data.json
{"id":1,"name":"zhangsan","age":"17","score":"75"}
{"id":2,"name":"lisi","age":"16","score":"80"}
{"id":3,"name":"wangwu","age":"18","score":"85"}
{"id":4,"name":"zhaoliu","age":"18","score":"90"}

$ cat data.json |jq '"id:\(.id),name:\(.name),age:\(.age),score:\(.score)"' -r
id:1,name:zhangsan,age:17,score:75
id:2,name:lisi,age:16,score:80
id:3,name:wangwu,age:18,score:85
id:4,name:zhaoliu,age:18,score:90

基础函数#

# has函数，检测对象是否包含key
$ echo -n '{"id":1,"name":"zhangsan","age":"17","score":"75"}'|jq 'has("id")'
true

# del函数，删除某个属性
$ echo -n '{"id":1,"name":"zhangsan","age":"17","score":"75"}'|jq 'del(.id)' -c
{"name":"zhangsan","age":"17","score":"75"}

# map函数，对数组中每个元素执行表达式计算，计算结果组织成新数组
$ seq 4|jq -s 'map(. * 2)' -c
[2,4,6,8]

# 上面map函数写法，其实等价于这个写法
$ seq 4|jq -s '[.[]|.*2]' -c
[2,4,6,8]

# keys函数，列出对象属性
$ echo -n '{"id":1,"name":"zhangsan","age":"17","score":"75"}'|jq 'keys' -c
["age","id","name","score"]

# to_entries函数，列出对象键值对
$ echo -n '{"id":1,"name":"zhangsan","age":"17","score":"75"}'|jq 'to_entries' -c
[{"key":"id","value":1},{"key":"name","value":"zhangsan"},{"key":"age","value":"17"},{"key":"score","value":"75"}]

# length函数，计算数组或字符串长度
$ jq -n '[1,2,3,4]|length'
4

# add函数，计算数组中数值之和
$ seq 4|jq -s 'add'
10

# tostring与tonumber，类型转换
$ seq 4|jq 'tostring|tonumber'
1
2
3
4

# type函数，获取元素类型
$ jq 'type' <<eof
1
"zhangsan"
true
null
{"id":1}
[75, 80, 85]
eof

"number"
"string"
"boolean"
"null"
"object"
"array"

过滤、排序、分组函数#

$ cat data.json
{"id":1,"name":"zhangsan","sex": 0, "age":"17","score":"75"}
{"id":2,"name":"lisi","sex": 1, "age":"16","score":"80"}
{"id":3,"name":"wangwu","sex": 0, "age":"18","score":"85"}
{"id":4,"name":"zhaoliu","sex": 0, "age":"18","score":"90"}

# select函数用于过滤，类似SQL中的where
$ cat data.json |jq 'select( (.id>1) and (.age|IN("16","17","18")) and (.name != "lisi") or (has("attr")|not) and (.score|tonumber >= 90) )' -c
{"id":3,"name":"wangwu","sex":0,"age":"18","score":"85"}
{"id":4,"name":"zhaoliu","sex":0,"age":"18","score":"90"}

# 有一些简化的过滤函数，如arrays, objects, iterables, booleans, numbers, normals, finites, strings, nulls, values, scalars
# 它们根据类型过滤，如objects过滤出对象，values过滤出非null值等
$ jq -c 'objects' <<eof
1
"zhangsan"
true
null
{"id":1}
[75, 80, 85]
eof

{"id":1}

$ jq -c 'values' <<eof
1
"zhangsan"
true
null
{"id":1}
[75, 80, 85]
eof

1
"zhangsan"
true
{"id":1}
[75,80,85]

# 选择出id与name字段，类似SQL中的select id,name
$ cat data.json|jq -s 'map({id,name})[]' -c
{"id":1,"name":"zhangsan"}
{"id":2,"name":"lisi"}
{"id":3,"name":"wangwu"}
{"id":4,"name":"zhaoliu"}

# 提取前2行，类似SQL中的limit 2
$ cat data.json|jq -s 'limit(2; map({id,name})[])' -c
{"id":1,"name":"zhangsan"}
{"id":2,"name":"lisi"}

# 按照age、id排序，类似SQL中的order by age,id
$ cat data.json|jq -s 'sort_by((.age|tonumber), .id)[]' -c
{"id":2,"name":"lisi","sex":1,"age":"16","score":"80"}
{"id":1,"name":"zhangsan","sex":0,"age":"17","score":"75"}
{"id":3,"name":"wangwu","sex":0,"age":"18","score":"85"}
{"id":4,"name":"zhaoliu","sex":0,"age":"18","score":"90"}


# 根据sex与age分组，并每组聚合计算count(*)、avg(score)、max(id)
$ cat data.json |jq -s 'group_by(.sex, .age)[]' -c
[{"id":1,"name":"zhangsan","sex":0,"age":"17","score":"75"}]
[{"id":3,"name":"wangwu","sex":0,"age":"18","score":"85"},{"id":4,"name":"zhaoliu","sex":0,"age":"18","score":"90"}]
[{"id":2,"name":"lisi","sex":1,"age":"16","score":"80"}]

$ cat data.json |jq -s 'group_by(.sex, .age)[]|{sex:.[0].sex, age:.[0].age, count:length, avg_score:map(.score|tonumber)|(add/length), scores:map(.score)|join(","), max_id:map(.id)|max }' -c                 
{"sex":0,"age":"17","count":1,"avg_score":75,"scores":"75","max_id":1}
{"sex":0,"age":"18","count":2,"avg_score":87.5,"scores":"85,90","max_id":4}
{"sex":1,"age":"16","count":1,"avg_score":80,"scores":"80","max_id":2}

字符串操作函数#

# contains函数，判断是否包含，实际也可用于判断数组是否包含某个元素
$ echo hello | jq -R 'contains("he")'
true

# 判断是否以he开头
$ echo hello | jq -R 'startswith("he")'
true

# 判断是否以llo结尾
$ echo hello | jq -R 'endswith("llo")'
true

# 去掉起始空格
$ echo ' hello '|jq -R 'ltrimstr(" ")|rtrimstr(" ")'
"hello"

# 大小写转换
$ echo hello|jq -R 'ascii_upcase'
"HELLO"

$ echo HELLO|jq -R 'ascii_downcase'
"hello"

# 字符串数组，通过逗号拼接成一个字符串
$ seq 4|jq -s 'map(tostring)|join(",")'
"1,2,3,4"

# json字符串转换为json对象
$ echo -n '{"id":1,"name":"zhangsan","age":"17","attr":"{\"weight\":56,\"height\":178}"}'|jq '.attr = (.attr|fromjson)' -c
{"id":1,"name":"zhangsan","age":"17","attr":{"weight":56,"height":178}}

# json对象转换为json字符串
$ echo -n '{"id":1,"name":"zhangsan","age":"17","attr":{"weight":56,"height":178}}'|jq '.attr = (.attr|tojson)'
{
  "id": 1,
  "name": "zhangsan",
  "age": "17",
  "attr": "{\"weight\":56,\"height\":178}"
}

$ cat data.txt
id:1,name:zhangsan,age:17,score:75
id:2,name:lisi,age:16,score:80
id:3,name:wangwu,age:18,score:85
id:4,name:zhaoliu,age:18,score:90

# 正则表达式过滤，jq使用的是PCRE
$ cat data.txt|jq -R 'select(test("id:\\d+,name:\\w+,age:\\d+,score:8\\d+"))' -r
id:2,name:lisi,age:16,score:80
id:3,name:wangwu,age:18,score:85

# 正则拆分字符串
$ cat data.txt|jq -R '[splits(",")]' -cr
["id:1","name:zhangsan","age:17","score:75"]
["id:2","name:lisi","age:16","score:80"]
["id:3","name:wangwu","age:18","score:85"]
["id:4","name:zhaoliu","age:18","score:90"]

# 正则替换字符串
$ cat data.txt |jq -R 'gsub("name"; "nick")' -r
id:1,nick:zhangsan,age:17,score:75
id:2,nick:lisi,age:16,score:80
id:3,nick:wangwu,age:18,score:85
id:4,nick:zhaoliu,age:18,score:90

# 正则表达式捕获数据
$ cat data.txt|jq -R 'match("id:(?<id>\\d+),name:(?<name>\\w+),age:\\d+,score:8\\d+")' -cr
{"offset":0,"length":30,"string":"id:2,name:lisi,age:16,score:80","captures":[{"offset":3,"length":1,"string":"2","name":"id"},{"offset":10,"length":4,"string":"lisi","name":"name"}]}
{"offset":0,"length":32,"string":"id:3,name:wangwu,age:18,score:85","captures":[{"offset":3,"length":1,"string":"3","name":"id"},{"offset":10,"length":6,"string":"wangwu","name":"name"}]}

# capture命名捕获，生成key是捕获组名称，value是捕获值的对象
$ cat data.txt|jq -R 'capture("id:(?<id>\\d+),name:(?<name>\\w+),age:\\d+,score:8\\d+")' -rc
{"id":"2","name":"lisi"}
{"id":"3","name":"wangwu"}

# 正则扫描输入字符串
$ cat data.txt|jq -R '[scan("\\w+:\\w+")]' -rc
["id:1","name:zhangsan","age:17","score:75"]
["id:2","name:lisi","age:16","score:80"]
["id:3","name:wangwu","age:18","score:85"]
["id:4","name:zhaoliu","age:18","score:90"]

日期函数#

# 当前时间缀
$ jq -n 'now'
1653820640.939947

# 将时间缀转换为0时区的分解时间(broken down time)，形式为 年 月 日 时 分 秒 dayOfWeek dayOfYear
$ jq -n 'now|gmtime' -c
[2022,4,29,10,45,5.466768980026245,0,148]

# 将时间缀转换为本地时区的分解时间(broken down time)
$ jq -n 'now|localtime' -c
[2022,4,29,18,46,5.386353015899658,0,148]

# 分解时间转换为时间串
$ jq -n 'now|localtime|strftime("%Y-%m-%dT%H:%M:%S")' -c
"2022-05-29T18:50:33"

# 与上面等效
$ jq -n 'now|strflocaltime("%Y-%m-%dT%H:%M:%SZ")'
"2022-05-29T19:00:40Z"

# 时间串解析为分解时间
$ date +%FT%T|jq -R 'strptime("%Y-%m-%dT%H:%M:%S")' -c
[2022,4,29,18,51,27,0,148]

# 分解时间转换为时间缀
$ date +%FT%T|jq -R 'strptime("%Y-%m-%dT%H:%M:%S")|mktime'
1653850310

高级用法#

实际上jq是一门脚本语言，它也支持变量、分支结构、循环结构与自定义函数，如下：

$ cat data.json
{"id":1,"name":"zhangsan","sex": 0, "age":"17","score":"75"}
{"id":2,"name":"lisi","sex": 1, "age":"16","score":"80"}
{"id":3,"name":"wangwu","sex": 0, "age":"18","score":"85"}
{"id":4,"name":"zhaoliu","sex": 0, "age":"18","score":"90"}

# 单变量定义
$ cat data.json| jq '.id as $id|$id'
1
2
3
4

# 对象展开式变量定义
$ cat data.json |jq '. as {id:$id,name:$name}|"id:\($id),name:\($name)"'
"id:1,name:zhangsan"
"id:2,name:lisi"
"id:3,name:wangwu"
"id:4,name:zhaoliu"

$ cat data.json
["1","zhangsan","17","75"]
["2","lisi","16","80"]
["3","wangwu","18","85"]
["4","zhaoliu","18","90"]

# 数组展开式变量定义
$ cat data.json|jq '. as [$id,$name]|"id:\($id),name:\($name)"'
"id:1,name:zhangsan"
"id:2,name:lisi"
"id:3,name:wangwu"
"id:4,name:zhaoliu"

# 分支结构
$ cat data.json|jq '. as [$id,$name]|if ($id>"1") then "id:\($id),name:\($name)" else empty end'
"id:2,name:lisi"
"id:3,name:wangwu"
"id:4,name:zhaoliu"

# 循环结构，第一个表达式条件满足时，执行只每二个表达式
# 循环结构除了while，还有until、recurse等
$ echo 1|jq 'while(.<100; .*2)'
1
2
4
8
16
32
64

# 自定义计算3次方的函数
$ echo 2|jq 'def cube: .*.*. ; cube'
8

由于这些高级特性并不常用，这里仅给出了一些简单示例，详细使用可以man jq查看。

辅助shell编程#

熟悉shell脚本编程的同学都知道，shell本身是没有提供Map、List这种数据结构的，这导致使用shell实现某些功能时，变得很棘手。

但jq本身是处理json的，而json中的对象就可等同于Map，json中的数组就可等同于List，如下：

list='[]';
#List添加元素
list=$(echo "$list"|jq '. + [ $val ]' --arg val java);
list=$(echo "$list"|jq '. + [ $val ]' --arg val shell);
#获取List大小
echo "$list"|jq '.|length'
#获取List第1个元素
echo "$list"|jq '.[0]' -r
# List是否包含java字符串
echo "$list"|jq 'any(.=="java")'
#删除List第1个元素
list=$(echo "$list"|jq 'del(.[0])');
# List合并
list=$(echo "$list"|jq '. + $val' --argjson val '["shell","python"]');
# List截取
echo "$list"|jq '.[1:3]'
# List遍历
for o in $(echo "$list" | jq -r '.[]');do 
    echo "$o"; 
done

map='{}';
#Map添加元素
map=$(echo "$map"|jq '.id=$val' --argjson val 1)
map=$(echo "$map"|jq '.courses=$val' --argjson val "$list")
#获取Map大小
echo "$map"|jq '.|length'
#获取Map指定key的值
echo "$map"|jq '.id' -r
#判断Map指定key是否存在
echo "$map" | jq 'has("id")'
#删除Map指定key
map=$(echo "$map"|jq 'del(.id)')
# Map合并
map=$(echo "$map"|jq '. + $val' --argjson val '{"code":"ID001","name":"hello"}')
# Map的KeySet遍历
for key in $(echo "$map" | jq -r 'keys[]'); do 
    value=$(jq '.[$a]' --arg a "$key" -r <<<"$map"); 
    printf "%s:%s\n" "$key" "$value"; 
done
# Map的entrySet遍历
while read -r line; do 
    key=$(jq '.key' -r <<<"$line"); 
    value=$(jq '.value' -r <<<"$line"); 
    printf "%s:%s\n" "$key" "$value"; 
done <<<$(echo "$map" | jq 'to_entries[]' -c)

总结#

可以发现，jq已经实现了json数据处理与分析的方方面面，我个人最近在工作中，也多次使用jq来分析调用日志等，用起来确实非常方便。

如果你现在还没完全学会jq的用法，没关系，建议先收藏起来，后面一定会用得到的！