@mattfarina Can you please update sprig to be compatible with the latest Masterminds/semver changes here: https://github.com/Masterminds/semver/pull/105?

Running go get -u github.com/Masterminds/sprig now returns the following error:

github.com/Masterminds/sprig ......\github.com\Masterminds\sprig\semver.go:22:2: cannot use semver.Version as type *semver.Version in return argument

semver.Version now needs to be a non-pointer, and semver_test.go needs to be updated to play nicely with this data type change.

We have golang applications that haven't migrated to go-modules yet, and that use sprig as a dependency. Can you please fix this compilation error so that our own downstream builds will succeed?

Thanks!

Eric

I have a case where I want a template to take an input dictionary and produce a copy, omitting a few keys at various levels deep in the hierarchy (in this helm chart).

I ran into a snare, where I didn't realize that after merging two dictionaries, editing the merged dictionary can modify the originals:

{{- $values := pick .Values "auth" "other"}}
{{- /* trim secret values. Update here if new secrets are added! */ -}}
{{- /* make a copy of values.auth to avoid modifying the original */ -}}
{{- $_ := set $values "auth" (merge dict .Values.auth) }}
{{- $_ := set $values.auth "state" (omit $values.auth.state "cryptoKey") }}

Without that set $values "auth" (merge dict .Values.auth), the subsequent set $values.auth "state" would modify the original .Values.auth.state, which I found surprising.

Is there an existing mechanism to produce a deep copy of a dict, to ensure the original doesn't get modified?

Closes Masterminds/sprig#95 (selfSignedCert cannot verify itself)

If it is intentional to leave the key usage out, we can discuss it here. Basically, let a self signed cert verify itself.

I'm using helm which depends on sprig. I wrote a simple fiddle which includes the helm template I'm testing with, as well as some go code that emulates how helm parses the template. You can see it here https://play.golang.org/p/imA_4tZJvPN

So I've got two objects, $app and $deployments, and if I run mergeOverwrite (dict) $app $deployments I would expect the new dict to first be mutated with the contents of $app and then with the contents of $deployments. But it seems that values from $deployments are being written into $app (as well as the new dict)

Without deepCopy(), Helm will be unusable for certain fairly simple tasks because of #120 : You cannot repeatedly use merge with the same "base" dict as the first argument unless you can tolerate that that dict will be modified every time. deepCopy() would avoid this.

Add float64 arithmetic functions. For clarity, these functions all have the "f" suffix. To avoid precision errors, all floats are converted to a decimal representation value using shopspring/decimal before calculating.

Given the following code:

{{ dict "foo" "bar" | merge (dict "foo" "abc") | toPrettyJson }}

The output is as expected:

{
  "foo": "abc"
}

But if the second dict's key is falsy, the value is not overwritten:

{{ dict "foo" "bar" | merge (dict "foo" 0) | toPrettyJson }}

Results in:

{
  "foo": "bar"
}

Any falsy value will be skipped:

{{ dict "foo" "bar" | merge (dict "foo" 0) | toPrettyJson }}
{{ dict "foo" "bar" | merge (dict "foo" "") | toPrettyJson }}
{{ dict "foo" "bar" | merge (dict "foo" false) | toPrettyJson }}
...etc

This is not the behavior of Javascript's Object.assign:

Object.assign({}, {foo: "bar"}, {foo: 0}) # <-- {foo: 0}

Sprig's behavior feels extremely counter-intuitive. When I'm merging dicts, if a key has a value assigned, that's the value I care about. It's language-specific falsyness feels irrelevant in this context.

Given:

{{- define "foobar" }}
foo:
  bar: baz
{{- end }}

qux:
{{- template "foobar" . | indent 2 }}

I'd like to get:

qux:
  foo:
    bar: baz

Given the above, I get:

executing "test" at <2>: wrong type for value; expected string; got map[string]interface {}

Using {{- template "foobar" . | toString | indent 2 }} I get:

qux:
foo:
  bar: baz

Meaning the line(s) never get indented. Even doing {{ template "foobar" . | join "_" | indent 2 }} results in the above, so clearly piping after template isn't working as I hoped it would.

Are there any ways to solve this?

This PR surfaces the random string functions that utilize crypto/rand that were added to goutils here: https://github.com/Masterminds/goutils/pull/25. This will allow Go templates to utilize random strings that have been generated in a cryptographically secure way.

I made the functions accessible by surfacing them through strings.go. I bounced back and forth between thinking they should be in strings.go and crypto.go a few times. If you think they should be made available via crypto.go instead, just let me know.

@technosophos

Thanks for building this lib! I really like it, but I'd like it even more if it haven't any dependency (outside the standard library)

Some of the current dependencies are to allow functions I don't care at all on templates, like UUID generation

For others, the following Go proverb applies:

"A little copying is better than a than a little dependency"

Rob Pike

For example, the xstrings package is imported just because of two functions

Proposal

Have a github.com/Masterminds/sprig/slim package that have all the functions that doesn't require importing external libraries

Hi guys!

This is a proposal to add a function to resolve names to IPs. I'm a heavy user of helm and this is useful where you need to render an IP but you don't want to hardcode it in your templates.

What do you think?

The ask is to have a function that acts like nindent, but skips the content's leading new line. The same effect is if the function act s as indent but doesn't indent the first line. This would allow templates to generate a leading new line that is useful in a lot of scenarios. The proposed name is snindent or sindent.

Explanation

A common way of using templates:

env:
  {{- include "my.multiline" . | nindent 2 }}
extra:
  field: abc

This assumes that "my.multiline" template produces a string without leading and trailing new lines. Let the template generate a list of "items" (line per item), each guarded by some condition.

{{- define "my.multiline" }}
{{- if condition1 }}
- line1
{{- end }}
{{- if condition2 }}
- line2
{{- end }}
{{- end }}

It is not possible to keep the two if blocks independent and keep no new lines around generated string for all possible values of boolean condition1 and condition2

So, the best for template is to generate a new line character per line: leading or trailing. If using trailing, then {{ include "sometemplate" . }} would generate likely unnecessary extra new line. {{ include "sometemplate" . -}} would in turn affect the next line indentation. So, a leading new line is the only option.

Workarounds

env:
  {{- include "my.multiline" . | trimPrefix "\n" | nindent 2 }}

This issue is ported over from helm issues but i am not sure if the issue I am seeing would be in https://github.com/imdario/mergo?

Output of helm version:

version.BuildInfo{Version:"v3.8.2", GitCommit:"6e3701edea09e5d55a8ca2aae03a68917630e91b", GitTreeState:"clean", GoVersion:"go1.18.1"}

Output of kubectl version:

N/A

Cloud Provider/Platform (AKS, GKE, Minikube etc.):

CAPI/Openstack, vanilla k8s.

Issue:

When doing a deepCopy, it is not possible to override an integer value with the value 0. I am not sure if this is a bug or a know behaviour (but it is a little annoying :-)).

We can reproduce this in the helm playground

With values:

---
rolloutStrategy1:
  type: RollingUpdate
  rollingUpdate:
    maxUnavailable: 1
    maxSurge: 1

rolloutStrategy2:
  type: RollingUpdate
  rollingUpdate:
    maxUnavailable: 0
    maxSurge: 1

and template:

---
{{- $nodeGroup := deepCopy $.Values.rolloutStrategy1 | mustMerge $.Values.rolloutStrategy2 }}
{{ toYaml $nodeGroup }}

I would expect the values in rolloutStrategy2 to override the values in rolloutStrategy1, which they do, unless the value is a 0.

Summary

It would be great to have a tar and gzip functions in order archive and compress string content to a tar.gz binary. This would mostly be useful to embed string content that is over the Kubernetes ConfigMap size limit of 1MBytes.

Proposal

The tar function would take as input a map[string][]byte where the key represents a file name and the value represents the file content.

The gzip function would take an array of bytes as input.

Usage Example

It could be used as follow in a Helm chart template:

my_compressed_files.tar.gz: |
  {{ .Files.Glob("foo/**") | tar | gzip }}

NOTE: The go standard library already provides a "archive/tar" package and a "compress/gzip" package.

Hello 👋 ,

I would like to use the really useful dig function within a range, but because the dig function only accepts "varargs" for path, I'm not able to use it.

With $someStruct equals to:

{
  "a": {
    "b": {
      "c": "value"
    }
  },
  "foo": {
    "bar": {
      "zoo": "value2"
    }
  }
}

First,

    {{- $keysToGet := list
      "foo.bar.zoo"
      "a.b.c"
    -}}
	{{ range $keysToGet }}
      {{ dig . "~" $someStruct }}
    {{ end }}

The result of this one is nil because dig looks for { "a.b.c": "value" } 🙂. So the result is twice ~ instead of value and value2.

Second, with array:

    {{- $keysToGet := list
      "foo.bar.zoo"
      "a.b.c.d.e"
    -}}
	{{ range $keysToGet }}
      {{ dig (splitList "." .) "~" $someStruct }}
    {{ end }}

Here, it's a type cast problem because the first parameter is not a string like expected by the function. I have the following error:

dig (splitList "." .) "~" $savedValues>: error calling dig: interface conversion: interface {} is []string, not string

It could be cool, especially in dynamic env like a range or with to have the ability to provide "path" coming from another variable dynamically

Thank you 😇

The current substr (substring) implementation does not respect runes (utf8 characters).

func main() {
	hello := "😞-Hello World-😞"
	fmt.Printf("substring = %s\n", substring(0, 5, hello))
	fmt.Printf("substringRune = %s\n", substringRune(0, 5, hello))
	fmt.Printf("substring = %s\n", substring(2, 7, hello))
	fmt.Printf("substringRune = %s\n", substringRune(2, 7, hello))
}

func substringRune(start, end int, s string) string {
	if start < 0 {
		return string([]rune(s)[:end])
	}
	if end < 0 || end > len(s) {
		return string([]rune(s)[start:])
	}
	return string([]rune(s)[start:end])
}

func substring(start, end int, s string) string {
	if start < 0 {
		return s[:end]
	}
	if end < 0 || end > len(s) {
		return s[start:]
	}
	return s[start:end]
}

Output substring = 😞- substringRune = 😞-Hel substring = ??-He substringRune = Hello

Remarks substring(): current implementation substringRune(): implementation that also works with runes

It would be useful if toDate and mustToDate functions accepted a parameter to specify the location for a time object that could be used to convert a time object between locations when UTC or the local location cannot be assumed.

For example, something like this I think would give you the UTC value for a date that was provided in : toDate "2006-01-02 15:04:05" "2017-12-31 01:01:01" "America/New_York" | date "2006-01-02T15:04:05Z"

Or could be used to convert a time received in Eastern to Pacific explicitly (I think I'm expressing this correctly): dateInZone "2006-01-02 15:04:05" (toDate "2006-01-02 15:04:05" "2017-12-31 01:01:01" "America/New_York") "America/Los_Angeles"