Incorrect JMESPath result with text output format

[jack-parsons-bjss]

Describe the bug

When the text output format is selected, results returned are inconsistent from the json output (and a test on jmespath.org, which is consistent with the json format).
We are running the following query: sort_by(imageDetails[? !(contains(imageTags[0], feature-branch))], & imagePushedAt)[-1].imageTags[0] on an aws ecr describe-images CLI call which ultimately results in more than one page of results being returned.

Regression Issue

• Select this option if this issue appears to be a regression.

Expected Behavior

I expect only the latest non feature branch version to be returned; in our case, 1.0.0-49

Current Behavior

I get that query executed on every page of results returned - 1.0.0-49 plus 1.0.0-18 by default, but by configuring the --page-size argument I can get more results.

Reproduction Steps

aws ec2 describe-images --owners amazon --query 'sort_by(Images[], & CreationDate)[-1].ImageId' --output text --page-size 100

Possible Solution

No response

Additional Information/Context

No response

CLI version used

aws-cli/2.17.42 Python/3.11.9 Darwin/23.6.0 source/arm64

Environment details (OS name and version, etc.)

MacOS Sonoma 14.7

[tim-finnigan]

Thanks for reaching out. Linking relevant AWS CLI v2 documentation for reference: https://docs.aws.amazon.com/cli/latest/userguide/cli-usage-output-format.html

As noted there, the difference in behavior between --output text and --output json is expected:

Regarding pagination, here is documentation with more information on how that works in the AWS CLI: https://docs.aws.amazon.com/cli/latest/userguide/cli-usage-pagination.html

When you specify --page-size for the describe-images command, that parameter is getting sent as MaxResults to the underlying API request. (You can see what the full request looks like when you add the --debug command.). Hope that helps, if you have any follow up questions please let us know.

[jack-parsons-bjss]

I understand that changing this behaviour is a breaking change and that you've said that this is expected, but from my perspective as a user this feels very unexpected. Why does changing the output format change the query behaviour rather than just changing how the output is rendered? I suppose I can close this as Won't Fix 🤷

[tim-finnigan]

Thanks for following up. I will close this as not planned as it is the expected behavior. I think this comment an older issue explains the reasoning behind the differences in behavior:

Generally the text output is for human readability. The JSON output is designed for scripting, we would suggest using either --query or pipe the JSON output to some other JSON parsing utility.

[github-actions]

This issue is now closed. Comments on closed issues are hard for our team to see.
If you need more assistance, please open a new issue that references this one.