[pkg/ottl] Decode function for decoding []byte based on a configured text encoding

[jsirianni]

Component(s)

pkg/ottl

Is your feature request related to a problem? Please describe.

Log receivers can output raw bytes (See Azure Event Hub). OTTL does not have the ability to parse these bytes into something human readable.

Describe the solution you'd like

A functionality similar to pkg/stanza/decode, which can take string or []byte input, and output it in a human readable format using a configured encoding (ascii, utf-8, etc).

Describe alternatives you've considered

• OTTL String function outputs Hexidecimal
• OTTL Base64Decode function is close (for the azure usecase) but cannot read []byte

Additional context

No response

[github-actions]

Pinging code owners:

• pkg/ottl: @TylerHelmuth @kentquirk @bogdandrutu @evan-bradley

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[TylerHelmuth]

We opted via #31543 to not make a generic Decode function as it would get bloated and there'd be edge cases. Which kind of decoding do you need to do?

[jsirianni]

We opted via #31543 to not make a generic Decode function as it would get bloated and there'd be edge cases. Which kind of decoding do you need to do?

The problem I am trying to solve starts with the Azure receiver outputting bytes. If you decode the output, you get human readable json that is perfectly usable. The problem is that OTEL does not offer the ability to decode []byte into anything useful.

Previously I asked to have the Azure Event Hub receiver output something human readable #32382, however it was explained that any receiver could output bytes, therefore the decode ability should live somewhere else.

[djaglowski]

We opted via #31543 to not make a generic Decode function as it would get bloated and there'd be edge cases. Which kind of decoding do you need to do?

Typically I'd agree on having targeted functions but I think in this case you're just basically moving a parameter into the function name and therefore causing bloat in the function set as well as friction while trying to add support for each encoding.

In pkg/stanza, the decode functionality has been pretty straightforward and stable. I don't see any discussion on #31543 regarding how bloat or edge cases would be problematic. Can you elaborate on this?

[TylerHelmuth]

My fear was that we'd end up needing to decode something that would take more parameters than just the encoding name. Maybe that fear is unwarranted? I see that Stanza supports these options, all as single parameters, which makes me feel better about adding a general Decode converter.

@djaglowski is there a reason stanza doesn't support base64 decoding in that list?

If my fear of edge-case parameters is off-base I'd support deprecating Base64Decode in favor of a generic function. The Converter has not existed for very long, so now is the time to do it.

/cc @evan-bradley

[djaglowski]

Thanks for considering @TylerHelmuth.

is there a reason stanza doesn't support base64 decoding in that list?

Not that I know of, just seems to have never been requested. I'm surprised it isn't listed in the golang.org/x/text/encoding/ianaindex list, but it would simple enough to add by just importing encoding/base64 and adding one more entry to the encodings map, "base64": base64.Encoding.

[TylerHelmuth]

I see that Go's implementation for base64 encode includes base64.StdEncoding, base64.URLEncoding, base64.RawStdEncoding, and base64.RawURLEncoding. The Base64Decode hardcodes base64.StdEncoding today, but could be made to specify which type via an optional parameter if that need ever came up.

We could also build that into a generic Decode function via base64, base64/url and so on.

[djaglowski]

We could also build that into a generic Decode function via base64, base64/url and so on.

Yeah, exactly. It's just a matter of articulating the strings and corresponding encoding.Encoding implementations which users will use to refer to each.

If my fear of edge-case parameters is off-base I'd support deprecating Base64Decode in favor of a generic function. The Converter has not existed for very long, so now is the time to do it.

I can't promise there won't ever be edge cases but the encoding.Encoding interface is very simple and robust for []byte => []byte and string => string conversions. I think in most cases if we find an unsupported encoding that we want to pick up, it'll just be a matter of adding it to the map.

[TylerHelmuth]

If @evan-bradley agrees, I am ok moving forward with this direction.

[evan-bradley]

I was mostly worried about adding complexity through per-encoding optional parameters. If we're reasonably confident we can keep Decode to a single parameter for the foreseeable future, I'm also okay with this.