@base64d does not handle web-safe encoded base64

[osresearch]

Describe the bug Many programs output "web safe" or "filename safe" base64 (defined in RFC 4648 and called base64url). In this format the / and + characters are replaced with _ and -, and the = padding is often URI escaped with %3D. The @base64d string formatter does not handle this variant.

To Reproduce

$ echo '_-8%3D' | jq -R -r '@base64d'
jq: error (at <stdin>:1): string ("_-%3D") is not valid base64 data

Expected behavior

$ echo '_-8%3D' | jq -R -r '@base64d' | xxd -g1
00000000: ff ef

Environment (please complete the following information):

• Ubuntu 20.04.
• jq-1.6

Additional context Having a way to output the raw binary would be helpful, too. The @base64d does not output the two bytes 0xFF 0xFE, but instead does a UTF-8 expansion.

[apstndb]

I define functions to work around.

def base64url: @base64 | gsub("\\+"; "-") | gsub("/"; "_") | gsub("="; "");
def base64urld: gsub("-"; "+") | gsub("_"; "/") | gsub("%3D"; "=") | @base64d;

Having a way to output the raw binary would be helpful, too. The @base64d does not output the two bytes 0xFF 0xFE, but instead does a UTF-8 expansion.

Binary supports seems working in this PR. https://github.com/stedolan/jq/pull/2314

[jfromaniello]

@apstndb thanks, this helped me to create a jwt-decode function:

~ » cat ~/.jq
def base64url: @base64 | gsub("\\+"; "-") | gsub("/"; "_") | gsub("="; "");
def base64urld: gsub("-"; "+") | gsub("_"; "/") | gsub("%3D"; "=") | @base64d;

~ » function jwt-decode () {
  jq -R 'split(".") | .[1] | base64urld | fromjson'
}