Hi Valdimir -
> Converting ECS to flat json where the fields take the form {"system.id":
"<value>"} is not a good option, because the very meaning of its use and
the convenience of the JSON format are lost.
Right, it just depends on your use case. My hope is that with the
facilities in Metron, you can manipulate the data in whatever manner works
best for you.
> And with deep nesting, this generally turns into unreadable,
hard-to-maintain code.
The overall approach with Metron has been to flatten nested fields, rather
than deal with deeply nested structures. Most of the parsers are built to
flatten the data. This presents a "flat" view for all downstream
functionality like Enrichment, Profiling, etc. I assume in your example
that whatever parser you are using is not flattening the data. Is that
correct?
> And now the question: is there a way to easily work with nested JSON in
Stellar? Deep diving into the documentation and source code has not yet
given an answer.
>
> system := MAP_PUT('id', MAP_GET('id',MAP_GET('os',host_id)), system)
>
Using a map literal would simplify your example a bit.
system := { 'id': MAP_GET('id', MAP_GET('os', host_id)) }
If we complete METRON-2072
<https://issues.apache.org/jira/browse/METRON-2072?jql=text%20~%20%22Stellar%20Map%22>,
adding some syntactic sugar around MAP_PUT/GET, then your example could be
much simpler.
system := { 'id': host_id['os']['id'] }
> Now this is a fundamentally important issue that affects the moments of
enrichment, TI, profiling and simply changing data when parsing.
All that being said, I think this highlights one advantage of using a DSL
like Stellar. If you do not want to flatten your data, it should be easy
enough to add whatever Stellar functions might be required to make the task
simpler.
I hope this helps.
On Fri, Nov 15, 2019 at 5:14 AM Vladimir Mikhailov <
[email protected]> wrote:
> Hi Nick!
>
> We, like Tom, plan to use Elastic Common Schema (ECS) to store events in
> Metron.
>
> A feature of ECS is the nesting of JSON objects, and therefore the "
> system.id" field implies storage in the form {"'system": {"id":
> "<value>"}}
>
> Converting ECS to flat json where the fields take the form {"system.id":
> "<value>"} is not a good option, because the very meaning of its use and
> the convenience of the JSON format are lost.
>
> Now, in order to work with nested JSON using Stellar, we are forced to use
> such complex constructs using the MAP_GET and MAP_PUT functions, for
> example:
>
> "fieldTransformations": [
> {
> "output": ["system"],
> "transformation": "STELLAR",
> "config": {
> "system": "MAP_PUT('id',
> MAP_GET('id',MAP_GET('os',host_id)), system)"
> }
> }
> ]
>
> And with deep nesting, this generally turns into unreadable,
> hard-to-maintain code.
>
> And now the question: is there a way to easily work with nested JSON in
> Stellar? Deep diving into the documentation and source code has not yet
> given an answer.
>
> Now this is a fundamentally important issue that affects the moments of
> enrichment, TI, profiling and simply changing data when parsing.
>
>
> On 2019/11/01 17:50:29, Nick Allen <[email protected]> wrote:
> > Hi Tom -
> >
> > > In the case of Metron, should we be modifying the field names to
> replace
> > dots? Can the Metron STELLAR language handle a field name with a dot in
> it,
> > or are there any special steps required such as surrounding event fields
> > with single or double-quotes in order to properly handle those field
> names?
> >
> > I cannot think of any facilities within Metron itself that would have
> > difficulties with periods in field names.
> >
> >
> > > I noticed that our fields names arrive with a period in the name, for
> > example "client.ip" and "user.id"... Our internal naming convention is
> > intended to align the data ingestion solution with the Elasticsearch
> Common
> > Schema. From experience, working with those dots in Elasticsearch is a
> > challenge
> >
> > You can use Metron to translate the field names however you like. For
> > example, replace "client.ip" with "client_ip". There are some examples of
> > this in the Parsers documentation here [1]
> > <
> https://metron.apache.org/current-book/metron-platform/metron-parsers/index.html
> >.
> > Looks under the section "fieldTransformation configuration".
> >
> >
> > ---
> >
> > [1]
> >
> https://metron.apache.org/current-book/metron-platform/metron-parsers/index.html
> >
> >
> >
> >
> >
> > On Fri, Nov 1, 2019 at 1:21 PM Yerex, Tom <[email protected]> wrote:
> >
> > > Good day to everyone. I'm working on our own variation of the
> Geographic
> > > Login Outliers use case (
> > >
> https://metron.apache.org/current-book/use-cases/geographic_login_outliers/index.html
> ).
> > > I noticed that our fields names arrive with a period in the name, for
> > > example "client.ip" and "user.id".
> > >
> > >
> > >
> > > Our internal naming convention is intended to align the data ingestion
> > > solution with the Elasticsearch Common Schema. From experience, working
> > > with those dots in Elasticsearch is a challenge and it raises the
> question
> > > if we need to handle field names with a dot in a different matter in
> Metron.
> > >
> > >
> > >
> > > In the case of Metron, should we be modifying the field names to
> replace
> > > dots? Can the Metron STELLAR language handle a field name with a dot
> in it,
> > > or are there any special steps required such as surrounding event
> fields
> > > with single or double-quotes in order to properly handle those field
> names?
> > >
> > >
> > >
> > > Thank you,
> > >
> > >
> > >
> > > Tom.
> > >
> > >
> > >
> >
>
