jq format when running from a bash script with variable expansion - bash

Question

I've got a jq command that works when running directly from the shell or from within a shell script, but when I try to add variable expansion, I get jq errors for unexpected format or invalid characters. My goal is to have a quick and easy way to update some json configuration.
Here's a simplified example.
The format of the json I'm modifying:
{
"pets": {
"some-new-pet": {
"PetInfo": {
"name": "my-brand-new-pet",
"toys": [
"toy1-postfix",
"toy2-postfix",
"toy3-postfix"
]
}
}
}
}
The jq without variable expansion:
cat myfile.json | jq '.pets."some-new-pet" += {PetInfo: {name: "my-brand-new-pet"}, toys: ["toy1", "toy2", "toy3"]}}'
The above runs fine, and adds the new pets.some-new-pet entry to my json.
Below is what I'm trying to do with variable expansion that fails.
jq_args = "'.pets.\"${PET}\" += {PetInfo: {name: \"${NAME}\"}, toys: [\"${toy1}-postfix\", \"${toy2}-postfix\", \"${toy3}-postfix\"]}}'"
cat myfile.json | jq $jq_args
The error message I get with the above:
jq: error: syntax error, unexpected INVALID_CHARACTER, expecting $end (Unix shell quoting issues?) at <top-level>, line 1: '.pets."some-new-pet"
My file is formatted as utf-8 and uses LF line endings.

Answer 1

I do not recommend constructing a jq filter using variable expansion or printf. It will work for simple cases but will fail if the string contains double quotes, backslashes or control-codes, as they have special meanings inside a JSON string. As an alternative to using printf, jq has a way to pass in variables directly via the command-line, avoiding all these issues.
pet='some-second-pet'
name='my-even-newer'
toy1=toy1
toy2=toy2
toy3=toy3
jq \
--arg pet "$pet" \
--arg name "$name" \
--arg toy1 "$toy1" \
--arg toy2 "$toy2" \
--arg toy3 "$toy3" \
'.pets[$pet] += {
PetInfo: {name: $name},
toys: ["\($toy1)-postfix", "\($toy2)-postfix", "\($toy3)-postfix"]
}' \
myfile.json
Output:
{
"pets": {
"some-new-pet": {
"PetInfo": {
"name": "my-brand-new-pet",
"toys": [
"toy1-postfix",
"toy2-postfix",
"toy3-postfix"
]
}
},
"some-second-pet": {
"PetInfo": {
"name": "my-even-newer-pet"
},
"toys": [
"toy1-postfix",
"toy2-postfix",
"toy3-postfix"
]
}
}
}

Answer 2

It would be cleaner and less error prone to format the string using printf
PET='dog'
NAME='sam'
toy1="t1"
toy2="t2"
toy3="t3"
jq_args=$(printf '.pets."%s" += {PetInfo: {name: "%s"}, toys: ["%s-postfix", "%s-postfix", "%s-postfix"]}}' "${PET}" "${NAME}" "${toy1}" "${toy2}" "${toy3}")
echo "$jq_args"
Result:
.pets."dog" += {PetInfo: {name: "sam"}, toys: ["t1-postfix", "t2-postfix", "t3-postfix"]}
Additionally, redundant quoting could be avoided by quoting the arg on this command
cat myfile.json | jq "$jq_args"

Answer 3

Fix your jq code by removing extra } at end
Fix bash jq call:
Add cotes "..." around your $jq_args
so don't use singles '...' in your jq_args definition
Use printf with -v option to define jq_args:
printf -v jq_args "...format..." value1 value2 ...
So your code became:
PET="some-new-pet"
NAME="my-brand-new-pet"
toy1="toy1"
toy2="toy2"
toy3="toy3"
format='.pets."%s" += {PetInfo: {name: "%s"}, toys: ["%s", "%s", "%s"]}'
printf -v jq_args "${format}" "${PET}" "${NAME}" "${toy1}" "${toy2}" "${toy3}"
cat myfile.json | jq "$jq_args"
Output:
{
"pets": {
"some-new-pet": {
"PetInfo": {
"name": "my-brand-new-pet"
},
"toys": [
"toy1",
"toy2",
"toy3"
]
}
}
}
Notes:
When you define your format, you put it into simple cotes '...'. It's really better to format JSON (or XML) without back-slashes (\\) before each double cotes (")
Use printf -v variable_name. It's more readable than var_name=$(printf ...)

Answer 4

By constructing the jq filter ("code") using outer bash variables ("data") you may run into escaping issues, which could eventually break or even divert your filter. (see https://en.wikipedia.org/wiki/Code_injection)
Instead, use mechanisms by jq to introduce external data through variables (parameter --arg):
jq --arg pet "${PET}" \
--arg name "${NAME}" \
--arg toy1 "${toy1}-postfix" \
--arg toy2 "${toy2}-postfix" \
--arg toy3 "${toy3}-postfix" \
'
.pets[$pet] += {PetInfo: {$name, toys: [$toy1,$toy2,$toy3]}}
' myfile.json
If you have an unknown number of variables to include, check out jq's --args parameter (note the additional s)