How to perform an replace-only update of a multidoc file

[BrendanGalloway]

For example, if I have a file:

a:
  b: 1
---
a:
  c: 2
---
a:
  b: 5

and want to update all existing instances of a.b (and only existing instances) to be 10, how would I do so?
if I use yq eval '.a.b = 10' the output is:

a:
  b: 10
---
a:
  c: 2
  b: 10
---
a:
  b: 10

which adds a new a.b entry, which is not desired. If I try use a filter such as yq eval '.a | select(has("b")) | .b = 10' then the output contains only the entries that matched the select, not the complete contents. And then only the child entries at that:

b: 10
---
b: 10

yq3 has a workaround in the merge command that lets me get the output I would like (even if it's a bit clunky to use):
yq m --overwrite --autocreate=false test.yml <(printf 'a:\n b: 10') -d=*

a:
  b: 10
---
a:
  c: 2
---
a:
  b: 10

is there an equivalent yq4 eval that gets the same output?

[mikefarah]

You just need to filter out the nodes that don't have .a.b on the LHS of the update operator:

yq e '(select(.a.b) | .a.b) = 10' examples/data1.yaml

[BrendanGalloway]

@mikefarah thanks, that works perfectly

[infa-ddeore]

@mikefarah how to replace all key, values? provided command updates the matching value, what i want is to replace a.b and a.c with provided a.b

input:

a:
  b: 1
---
a:
  c: 2
---
a:
  b: 5
  e: 3

$ cat /tmp/4 | yq4 '(select(.a.b) | .a.b) = 20'
a:
  b: 20
---
a:
  c: 2
---
a:
  b: 20
  e: 3

expected output

a:
  b: 20
---
a:
  c: 2
---
a:
  b: 20

[mikefarah]

I'm confused - what are you replacing with what? the expected output, a.c hasn't changed, but the third doc a has lost e and has had b updated... not sure what the intention is here?

[infa-ddeore]

@mikefarah, i want to replace all elements, expected output is as below where in last element a.e is removed and contains only given value(s)

a:
  b: 20
---
a:
  c: 2
---
a: 
  b: 20