Splunk Search

How to use spath to read dynamic path

mayurkale471757
Explorer

Hi Team I have the below Json string coming as an event in Splunk logs .

after data, the next field could be a, b, c, d 

I want to read the x and y fields, How to write a single spath query like 

| spath input=inputJson  path="data.{*}.x" 

 

 

{data : {a : {  x: {  }  y: {  }}} }

{data : {b : {  x: { }  y: {  }}} }

{data : {c : {  x: {  }  y: {  }}} }

{data : {d : {  x: {  }  y: {  }}} }

 

 

 

Labels (3)
Tags (1)
0 Karma
1 Solution

yuanliu
SplunkTrust
SplunkTrust

You need to first extract data beyond the "dynamic" key. (Depending on semantics, I suspect that there is some data design improvement your developers could make so downstream users don't have to do this goaround.)

 

 

| spath input=json_data path=data output=beyond
| eval key = json_array_to_mv(json_keys(beyond))
| eval beyond = json_extract(beyond, key) ``` assuming there is only one top key ```
| spath input=beyond path=x
| spath input=beyond path=y

 

 

The following is full emulation (I don't see the purpose of all the transposes)

 

| makeresults count=1
| eval json_data="{\"data\": {\"a\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"b\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"c\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"d\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| spath input=json_data path=data output=beyond
| eval key = json_array_to_mv(json_keys(beyond))
| eval beyond = json_extract(beyond, key)
| spath input=beyond path=x
| spath input=beyond path=y
| table json_data x y beyond

 

json_dataxybeyond
{"data": {"a": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}
{"data": {"b": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}
{"data": {"c": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}
{"data": {"d": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}

View solution in original post

efavreau
Motivator

Note:
1) The spath command can be expensive, especially against large data sets
2) If all you need is to parse a string and get the values, consider regular expressions for json data also.

In the rex below, I named the a|b|c|d field "foo", in case it had value later on. If not, it doesn't need to be used

| makeresults ```creating dummy data based on the original question```
| eval json_data="{data: {a : {  x: {value_x}  y: {value_y}}} }" 
| append 
    [ makeresults
    | eval json_data="{data: {b : {  x: {value_x}  y: {value_y}}} }"
        ] 
| append 
    [ makeresults
    | eval json_data="{data: {c : {  x: {value_x}  y: {value_y}}} }"
        ] 
| append 
    [ makeresults
    | eval json_data="{data: {d : {  x: {value_x}  y: {value_y}}} }"
        ]
```ending the creation of dummy data```
| rex field=json_data "{(?<foo>\w+)\s:\s{\s\sx:\s{(?<x_value>.+)}\s\sy:\s{(?<y_value>.+)}}}" ```parse strings using a regular expression```
| table json_data x_value y_value ```display results of regular expression in a table```

Results in:

efavreau_1-1707149703942.png

 

###

If this reply helps you, an upvote would be appreciated.
0 Karma

PickleRick
SplunkTrust
SplunkTrust

There is no syntax for this. With spath you have to either provide a precise path or not provide path at all so the whole source field (_raw by default) is parsed.

You can later do some magic based most probably on foreach and matching field name to a pattern but that's kinda ugly and not very efficient.

Anyway, it seems like a badly designed data schema because it looks as if you should rather have an array od objects instead of different objects. Conceptually - different objects are different types of entity so why would you want to treat them the same?

0 Karma

ITWhisperer
SplunkTrust
SplunkTrust

It depends what you want to do next - if you just want to remove the a, b, c, and d from the field names, you could just do this

| spath
| fields - _raw
| transpose 0 column_name=field
| eval field=mvindex(split(field,"."),0).".".mvindex(split(field,"."),2).".".mvindex(split(field,"."),3)
| transpose 0 header_field=field
| fields - column
0 Karma

mayurkale471757
Explorer

It is producting below result , 

I want to read the x and y field

 

| makeresults count=1
| eval json_data="{\"data\": {\"a\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"b\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"c\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"d\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| spath
| fields - _raw
| transpose 0 column_name=field
| eval field=mvindex(split(field,"."),0).".".mvindex(split(field,"."),2).".".mvindex(split(field,"."),3)
| transpose 0 header_field=field
| fields - column

 

mayurkale471757_0-1706866358989.png

 

0 Karma

yuanliu
SplunkTrust
SplunkTrust

You need to first extract data beyond the "dynamic" key. (Depending on semantics, I suspect that there is some data design improvement your developers could make so downstream users don't have to do this goaround.)

 

 

| spath input=json_data path=data output=beyond
| eval key = json_array_to_mv(json_keys(beyond))
| eval beyond = json_extract(beyond, key) ``` assuming there is only one top key ```
| spath input=beyond path=x
| spath input=beyond path=y

 

 

The following is full emulation (I don't see the purpose of all the transposes)

 

| makeresults count=1
| eval json_data="{\"data\": {\"a\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"b\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"c\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"d\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| spath input=json_data path=data output=beyond
| eval key = json_array_to_mv(json_keys(beyond))
| eval beyond = json_extract(beyond, key)
| spath input=beyond path=x
| spath input=beyond path=y
| table json_data x y beyond

 

json_dataxybeyond
{"data": {"a": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}
{"data": {"b": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}
{"data": {"c": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}
{"data": {"d": {"x": {"mock_x_field": "value_x"}, "y": {"mock_y_field": "value_y"}}}}{"mock_x_field":"value_x"}{"mock_y_field":"value_y"}{"x":{"mock_x_field":"value_x"},"y":{"mock_y_field":"value_y"}}

mayurkale471757
Explorer

Thanks @yuanliu  this is what I was looking for great !!

0 Karma

ITWhisperer
SplunkTrust
SplunkTrust
| makeresults count=1
| eval json_data="{\"data\": {\"a\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"b\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"c\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| append [ makeresults count=1
          | eval json_data="{\"data\": {\"d\": {\"x\": {\"mock_x_field\": \"value_x\"}, \"y\": {\"mock_y_field\": \"value_y\"}}}}"
        ]
| spath input=json_data
| fields - json_data
| transpose 0 column_name=field header_field=_time
| eval field=mvindex(split(field,"."),0).".".mvindex(split(field,"."),2).".".mvindex(split(field,"."),3)
| transpose 0 header_field=field
| fields - column
0 Karma
Get Updates on the Splunk Community!

How to Monitor Google Kubernetes Engine (GKE)

We’ve looked at how to integrate Kubernetes environments with Splunk Observability Cloud, but what about ...

Index This | How can you make 45 using only 4?

October 2024 Edition Hayyy Splunk Education Enthusiasts and the Eternally Curious!  We’re back with this ...

Splunk Education Goes to Washington | Splunk GovSummit 2024

If you’re in the Washington, D.C. area, this is your opportunity to take your career and Splunk skills to the ...