nifi-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joey Frazee <joey.fra...@icloud.com>
Subject Re: Enrichment plugin for adding attributes from SQL
Date Sun, 07 Jan 2018 20:58:05 GMT
Hey Brett,

So as for the LookupAttribute only doing a single key lookup, that was a bit of a coin flip
on whether it makes more sense to allow multiple lookups at once, each with a single key,
or a single lookup with multiple AND-ed constraints. I opted for the former since that was
the use case I was going after and we don’t really have a principled way of doing a map
or set-valued property, so we do multiple properties and pretend it’s a map.

I think I considered enabling a flag to choose between the two alternatives, so that’d be
the easy, dumb thing I could do and then we could relax those requirements (which are correct
to have at the moment). I think the difficult part would really be more in writing a custom
validator — and that wouldn’t be hard either.

-joey

On Jan 7, 2018, 5:23 AM -0600, Brett Ryan <brett.ryan@gmail.com>, wrote:
> MongoDBLookupService can't be used with an UpdateAttribute processor though, it returns
a Record type, and has no required keys. The whole purpose of the processor that I originally
wrote was to update an attribute from SQL on a FF.
>
> The implementation of LookupAttribute requires
>
> - getRequiredKeys to return a single element set
> - getValueType to return String.class
>
> I guess what I could do is
>
> (1) Write a processor that updates an attribute on a FF from a Record based LookupService.
Say I call it UpdateAttributeFromRecordLookup
> (2) Write a service SQLLookupService which given a query as a lookup coordinate would
return a Record for the first found entry.
> (3) Write a processor that updates an attribute from a record lookup, call it UpdateAttributeFromRecordLookup.
>
> It's a bit more long winded to achieve what I originally had with a single processor,
but I guess UpdateAttributeFromRecordLookup would have the benefit of updating attributes
from LookupService implementations that do not meet the LookupAttribute criteria.
>
> If you think this is viable, please let me know and I'll have a look at starting it tomorrow
afternoon.
>
>
> > On 7 Jan 2018, at 03:23, Mike Thomsen <mikerthomsen@gmail.com> wrote:
> >
> > Take a look at the mongo lookup service. I think it could serve as a good
> > example here.
> > On Fri, Jan 5, 2018 at 10:49 PM Brett Ryan <brett.ryan@gmail.com> wrote:
> >
> > > Looking at using a lookupservice, this doesn't seem to support sending
> > > multiple keys to the LookupService at the same time.
> > >
> > > What I was thinking of doing was implement a LookupService that took an
> > > attribute "sql.query" which would use this to evaluate the query but then
> > > pass in a map of key/value pairs for attribute-name/column-name to set the
> > > attributes.
> > >
> > > I could implement this as I imagined it to work, however it will evaluate
> > > the SQL expression multiple times for the same query on the same flow.
> > >
> > > I am also wondering why LookupService#getRequiredKeys must return a
> > > Set<String>, yet; this set must only contain one value.
> > >
> > >
> > > On 5 Jan 2018, at 08:32, Andy LoPresto <alopresto@apache.org> wrote:
> > >
> > > UpdateAttribute doesn’t pull from a database, it uses static or dynamic
> > > attribute values and supports NiFi Expression Language. In your original
> > > message, you didn’t mention any database interaction, so I thought you were
> > > just trying to accomplish "I wanted to add some attributes to a FlowFile
> > > while not altering the contents of that FlowFile”, which is indeed what
> > > UpdateAttribute does.
> > >
> > > If you need to retrieve those values from a database, as Joey mentions,
> > > the LookupService is the right tool.
> > >
> > > With your prior setup, the distributed map cache is as secure as the NiFi
> > > configuration — if using secured NiFi, the communication between that node
> > > and any other is over TLS, and within the node it’s a memory access.
> > >
> > > A big part of the NiFi philosophy is the same as the Unix philosophy —
> > > each tool should do one job very well, and to perform complicated tasks,
> > > chain those tools together. This helps with provenance reporting, usage
> > > reporting, debugging, flow development lifecycle, maintenance, etc. A
> > > processor which retrieves attributes from a database and updates the
> > > incoming flowfile with them is certainly useful in the use case you
> > > describe, but is not a generic pattern. There’s no intent to discourage
> > > custom development, and whatever makes your flow work is fine. Just
> > > explaining why you likely won’t see a solution like that in the NiFi
> > > bundled components.
> > >
> > >
> > > Andy LoPresto
> > > alopresto@apache.org
> > > *alopresto.apache@gmail.com <alopresto.apache@gmail.com>*
> > > PGP Fingerprint: 70EC B3E5 98A6 5A3F D3C4  BACE 3C6E F65B 2F7D EF69
> > >
> > > On Jan 4, 2018, at 4:07 PM, Brett Ryan <brett.ryan@gmail.com> wrote:
> > >
> > > Thanks Andy, how would update attribute be able to get the value from sql?
> > >
> > > Consider a flow where a piece of information needs to be obtained from a
> > > DB but i do not want the contents of the current FF to be altered, using
> > > ExecuteSQL anywhere prior would not be possible due to replacing the FF
> > > contents.
> > >
> > > What I had was two seperate flows, one that updates an oauth key in a db
> > > keeping it fresh, the main flow would then read the db just before an
> > > invokehttp.
> > >
> > > I was originally using a distributed map cache but had concerns that it
> > > might not be secure and was also advised the cache server has been known to
> > > go down.
> > >
> > > On 5 Jan 2018, at 04:01, Joey Frazee <joey.frazee@icloud.com> wrote:
> > >
> > > Andy, Brett,
> > >
> > > Taking a quick glance at the code it looks like it's enriching attributes
> > > from a database according to a query.
> > >
> > > If that's correct, there's a LookupAttribute processor that delegates
> > > lookups to a "LookupService" and adds attributes without altering content.
> > > There are a variety of these LookupServices included. I think what you
> > > implemented would make sense as a LookupService and then you could just
> > > configure the processor to use that. It could also be used with
> > > LookupRecord then too so there'd be a double payoff.
> > >
> > > -joey
> > >
> > > On Jan 4, 2018, 8:44 AM -0800, Andy LoPresto <alopresto@apache.org>,
> > > wrote:
> > > Hi Brett,
> > >
> > > It’s great that you found it easy to write a new processor for Apache
> > > NiFi. It is probably an indicator that we need to improve
> > > education/evangelism/documentation, however, that you did not find
> > > UpdateAttribute [1], which should do exactly what you were looking for.
> > >
> > > [1]
> > > https://nifi.apache.org/docs/nifi-docs/components/org.apache.nifi/nifi-update-attribute-nar/1.4.0/org.apache.nifi.processors.attributes.UpdateAttribute/index.html
> > >
> > > Andy LoPresto
> > > alopresto@apache.org
> > > alopresto.apache@gmail.com
> > > PGP Fingerprint: 70EC B3E5 98A6 5A3F D3C4  BACE 3C6E F65B 2F7D EF69
> > >
> > > On Jan 4, 2018, at 7:03 AM, Brett Ryan <brett.ryan@gmail.com> wrote:
> > >
> > > Hi all, having used NiFi for a couple days I wanted to add some attributes
> > > to a FlowFile while not altering the contents of that FlowFile.
> > >
> > > I had suggestions to use a script processor but that just sounded like a
> > > hack which could become a nuisance to replicate.
> > >
> > > Anyway, I figured I'd write a processor to do this, anyone interested you
> > > can find it here.
> > >
> > > https://github.com/brettryan/nifi-drunken-bundle
> > >
> > > Feel free to do with it what you please.
> > >
> > > I've published to maven central but it will take a day to appear in the
> > > search.
> > >
> > > <dependency>
> > > <groupId>com.drunkendev</groupId>
> > > <artifactId>nifi-drunken-nar</artifactId>
> > > <version>1.0.0</version>
> > > <type>nar</type>
> > > </dependency>
> > > <dependency>
> > > <groupId>com.drunkendev</groupId>
> > > <artifactId>nifi-drunken-processors</artifactId>
> > > <version>1.0.0</version>
> > > </dependency>
> > > <dependency>
> > > <groupId>com.drunkendev</groupId>
> > > <artifactId>nifi-drunken-bundle</artifactId>
> > > <version>1.0.0</version>
> > > <type>pom</type>
> > > </dependency>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message