Remove subelement from a yaml array in bash/awk when there's no order? - awk

Question

I'm trying to find a better way of removing values from a yaml, for example - this is my yaml example:
apiVersion: v1
data:
mapRoles: |-
- username: user1
rolearn: arn
groups:
- grp
- grp2
- groups:
- grp
rolearn: arn
username: user2
kind: ConfigMap
metadata:
annotations:
kubectl.kubernetes.io/last-applied-configuration: |
{"..."}
uid: 93ad6dc1-2a1f-11ea-b5da-0ec0e91c7076
my input is list of user names, which i can test with regex or. as i cannot install any dependencies, i have to use a tool that is installed in any system - thats what i chose awk.
in each part, i have to check the username if it matches any list of values, then if it does - remove a specific group from the "groups:" list.
what i was thinking is to identify each start of a yaml key (that represents a user) - then, add everything to an array while checking if the username is exactly what we expect - if it does, print the array but without the relevant group, else - print the entire array.
i've started writing it and it seems complex - is there any better way?
--- examples ---
If i'm specifying "user1" and the "grp" as the params, yaml should look like:
apiVersion: v1
data:
mapRoles: |-
- username: user1
rolearn: arn
groups:
- grp2
- groups:
- grp
rolearn: arn
username: user2
kind: ConfigMap
metadata:
annotations:
kubectl.kubernetes.io/last-applied-configuration: |
{"..."}
uid: 93ad6dc1-2a1f-11ea-b5da-0ec0e91c7076
if i'm specifying user2 and the "grp", it should look like:
apiVersion: v1
data:
mapRoles: |-
- username: user1
rolearn: arn
groups:
- grp
- grp2
- groups:
rolearn: arn
username: user2
kind: ConfigMap
metadata:
annotations:
kubectl.kubernetes.io/last-applied-configuration: |
{"..."}
uid: 93ad6dc1-2a1f-11ea-b5da-0ec0e91c7076
That's my issue - basically the user2 is specified AFTER the groups section, which then i'm not sure about the correct way to remove it.

Answer 1

This might be what you're trying to do but its not clear from your question:
$ cat tst.awk
BEGIN {
split(users,tmp)
for (i in tmp) {
tgtUsers[tmp[i]]
}
split(groups,tmp)
for (i in tmp) {
tgtGroups[tmp[i]]
}
}
match($0,/^[[:space:]]*(-[[:space:]]*)?[^[:space:]]+[[:space:]]*:/) {
sect = $0
sub(/^[[:space:]]*(-[[:space:]]*)?/,"",sect)
sub(/[[:space:]]*:.*/,"",sect)
}
sect == "username" {
inTgtUsers = ($NF in tgtUsers)
inGroups = 0
}
sect == "groups" {
inGroups = 1
}
!(inGroups && inTgtUsers && ($NF in tgtGroups))
$ awk -v users='user1' -v groups='grp' -f tst.awk file
apiVersion: v1
data:
mapRoles: |-
- username: user1
rolearn: arn
groups:
- grp2
- groups:
rolearn: arn
username: user2
kind: ConfigMap
metadata:
annotations:
kubectl.kubernetes.io/last-applied-configuration: |
{"..."}
uid: 93ad6dc1-2a1f-11ea-b5da-0ec0e91c7076

Answer 2

So, by the note of Ed Morton, i've taken his script, but had to change the regex, as it didn't catch groups with colons in the name (such as - my:group)
BEGIN {
inGroups = 0
split(users,tmp)
for (i in tmp) {
tgtUsers[tmp[i]]
}
split(groups,tmp)
for (i in tmp) {
tgtGroups[tmp[i]]
}
}
match($0,/^ +(-? (username|rolearn|userarn): [^ ]+|-? groups: *$)/) {
sect = $0
sub(/^[[:space:]]*(-[[:space:]]*)?/,"",sect)
sub(/[[:space:]]*:.*/,"",sect)
if ($1 == "-") {
startIndexArr[NR]
start_index = NR
}
inGroups = 0
}
sect == "username" {
if ($NF in tgtUsers)
relevantUsersIndexArr[start_index]
}
sect == "groups" {
inGroups = 1
}
(inGroups == 1 && ($NF in tgtGroups)) {
foundGroupsArr[NR]
}
{
yamlArr[NR] = $0
}
END {
row_num = 1
for (i in yamlArr) {
if (row_num in startIndexArr)
start_entity_index = row_num
print_row = 1
if (row_num in foundGroupsArr && start_entity_index in relevantUsersIndexArr)
print_row = 0
if (print_row == 1)
print yamlArr[row_num]
row_num++
}
}