v3.0.0

[TomWright]

This has been in the works for a long time, and is almost a complete rewrite of dasel.

A lot of features have been added and I'm sure there will be some issues - please let me know about these and I'll fix them.

See the updated documentation.

Added

• Major new version release.
• INI support.
• HCL support.
• Dasel syntax now supports variables and expressions.
• Files can now be read and parsed inside a dasel query.
• Variables can now be passed to dasel from the command line.
• Support for comments in queries.
• Dasel config file to define default file format.
• Interactive mode for dasel CLI (alpha).

Changed

• Go module path changed to github.com/tomwright/dasel/v3.
• Internal changes to support new version.
• Query/selector syntax revamp. See docs for more information.
• Majority of read/write operations will now maintain ordering.
• Migrated from Cobra to Kong for CLI parsing/processing.
• Removed put and delete commands. Instead, modify within the query and use --root flag.

Fixed

• File redirect now works in the same way as piped input.
• Various other bug fixes and improvements.
• Whitespace in query syntax is now handled correctly.

---

This discussion was created from the release v3.0.0.

[pmeier]

@TomWright Thanks a ton for the update, I really like the new query syntax! Good job!

I have an issue though replacing dasel put from v2. I cannot find a way to do inplace updates. For example, I used dasel to inject dynamically computed versions into a helm chart:

dasel put --file ${CHART} version --value 0.0.0
dasel put --file ${CHART} appVersion --value 0.0.0
cat ${CHART}

What I have for v3 now is

NEW_CHART=$(dasel --var="chart=yaml:file:${CHART}" --out=yaml \
    '{$chart..., "version": "0.0.0", "appVersion": "0.0.0"}'  \
)
echo "${NEW_CHART}" | tee ${CHART}

Unfortunately, I cannot dasel --var=...:$CHART | tee $CHART. Is there a way to specify an output file or do I have to write it manually?

[TomWright]

Thanks for the question.

Good call out on the in-place updates. There is some info on this here, although it is somewhat hidden.

As for dasel handling the update in-place, I have chosen to remove file writes as it feels like a separate issue.

Here's a sample example of in-place edits that will work:

➜  ~ cat Chart.yaml
apiVersion: v1
kind: ConfigMap
metadata:
    name: mychart-configmap
data:
    myvalue: Hello World

➜  ~ NAME="newName" A="foo" B="bar" dasel -i yaml --root 'metadata = {
    metadata...,
    foo: $A,
    bar: $B,
    name: $NAME
}' < Chart.yaml > Chart.yaml.tmp \
&& mv Chart.yaml.tmp Chart.yaml

➜  ~ cat Chart.yaml
apiVersion: v1
kind: ConfigMap
metadata:
    name: newName
    foo: foo
    bar: bar
data:
    myvalue: Hello World

I'll add some examples to the documentation for this use-case. Thank you!

[TomWright]

As an aside, if this does become a recurring issue for many users I can re-implement to simplify this flow

[arturmartins]

Hello. I presume that the migration from Cobra to Kong explains why the completion command is no longer available?

[TomWright]

That is correct, although I have been looking at Kong addons to do the same thing.

This is near the top of my todo list regarding features to add.

[electriquo]

The docs and discussions used to hold many examples, but due to the rewrite and the major updates, seems like not everything is backward compatible.

1. What do you think of adding examples for clarity?
2. CSV does not seem to work? For instance, I am tying to query for all the values of a specific column\header (I tried all kinds of queries):

   $ echo '
   col1,col2
   a,aa
   b,bb
   c,cc
   ' | dasel -i csv '.col1'
   dasel: error: error parsing selector: failed to parse: unexpected token 29 "." at position 0.

[TomWright]

I can add some examples for sure. Sorry it was a large rewrite and I honestly lost track of which examples were missing.

Here's an example of trimming the data down:

echo '
col1,col2
a,aa
b,bb
c,cc
' | dasel -i csv -o json 'map(col2)'
[
    "aa",
    "bb",
    "cc"
]

Or:

$ echo '
col1,col2
a,aa
b,bb
c,cc
' | dasel -i csv -o json 'map(col2)...'
"aa"

"bb"

"cc"

The CSV writer expects an array of objects. col2 returns an array of strings which confuses it.

If you want a CSV with a single column you can also do this:

$ echo '
col1,col2
a,aa
b,bb
c,cc
' | dasel -i csv 'map({value: col1})'
value
a
b
c

Or to extract a subset of the columns, simply:

$ echo '
col1,col2
a,aa
b,bb
c,cc
' | dasel -i csv 'map({col1})'
col1
a
b
c

If you have an exact output you're after let me know and I'll provide an example.

Once the holiday period is over I'll attempt to add missing examples to the documentation. Thanks for your patience and input!

[electriquo]

@TomWright thank you for your amazing work in creating the community, keeping it alive, and publishing the tool. It's deeply appreciated!

You may have noticed that I needed to be very active when using dasel due to ongoing development, (breaking) changes, and the lack of documentation that fully showcases its capabilities. As I mentioned to you before, I eventually decided it would be more effective for me to invest my time in developing my own tool that supports all file formats.

The core idea was simple: determine the input file format, convert it into a native/primitive object, and then apply a query to that object — where the query can use the language's native syntax. In implementations like NodeJS/Python/Ruby, the input becomes an Object/Dictionary/Hash (respectively), and the query is applied directly. This means the query syntax, results, errors, and documentation are all those of the underlying programming language are well-defined, maintained, and familiar.

I'm sharing this again as food for thought, and once more thank you!