[DISCUSS] Spargel vs. Gelly

[Ufuk Celebi-2]

Hey all,

currently we have two vertex-centric graph APIs: Spargel and Gelly. I want to discuss whether we shall
1) keep Spargel as a public API as it is, or
2) deprecate (and remove) it.

In my understanding, Spargel was a proof-of-concept, which stuck around. It is very stable, but limited in functionality. Gelly provides a superset of Spargel's functionality and a high-level library of graph algorithms. The vertex-centric iterations actually wrap Spargel.

I am in favour of 2):

+ Less confusing and less work to have two APIs for the same thing (we have to communicate this, document it etc.)
+ Gelly is actively maintained and getting a lot of contributions
- Spargel users will have to move to Gelly at some point in time (I think this will happen anyways and it should be straight forward)

The Spargel internal code will probably stick around as part of Gelly. The question is whether we want to have two public APIs for this.

– Ufuk

[Andra Lungu]

+1 for removing Spargel, but you have to admit that my opinion is pretty
biased :D

On Fri, Apr 10, 2015 at 12:13 PM, Ufuk Celebi <[hidden email]> wrote:

> Hey all,
>
> currently we have two vertex-centric graph APIs: Spargel and Gelly. I want
> to discuss whether we shall
> 1) keep Spargel as a public API as it is, or
> 2) deprecate (and remove) it.
>
> In my understanding, Spargel was a proof-of-concept, which stuck around.
> It is very stable, but limited in functionality. Gelly provides a superset
> of Spargel's functionality and a high-level library of graph algorithms.
> The vertex-centric iterations actually wrap Spargel.
>
> I am in favour of 2):
>
> + Less confusing and less work to have two APIs for the same thing (we
> have to communicate this, document it etc.)
> + Gelly is actively maintained and getting a lot of contributions
> - Spargel users will have to move to Gelly at some point in time (I think
> this will happen anyways and it should be straight forward)
>
> The Spargel internal code will probably stick around as part of Gelly. The
> question is whether we want to have two public APIs for this.
>
> – Ufuk

[Stephan Ewen]

I would remove Spargel at some point in time, but not until Gelly is out of
staging.

On Fri, Apr 10, 2015 at 12:13 PM, Ufuk Celebi <[hidden email]> wrote:

> Hey all,
>
> currently we have two vertex-centric graph APIs: Spargel and Gelly. I want
> to discuss whether we shall
> 1) keep Spargel as a public API as it is, or
> 2) deprecate (and remove) it.
>
> In my understanding, Spargel was a proof-of-concept, which stuck around.
> It is very stable, but limited in functionality. Gelly provides a superset
> of Spargel's functionality and a high-level library of graph algorithms.
> The vertex-centric iterations actually wrap Spargel.
>
> I am in favour of 2):
>
> + Less confusing and less work to have two APIs for the same thing (we
> have to communicate this, document it etc.)
> + Gelly is actively maintained and getting a lot of contributions
> - Spargel users will have to move to Gelly at some point in time (I think
> this will happen anyways and it should be straight forward)
>
> The Spargel internal code will probably stick around as part of Gelly. The
> question is whether we want to have two public APIs for this.
>
> – Ufuk

[Vasiliki Kalavri]

Hey!

We've had this discussion before actually [1] and we also have a JIRA to
deprecate Spargel [2] :-)
I agree that we should keep Spargel deprecated until Gelly is out of
staging.

Cheers,
-Vasia.

[1]: https://www.mail-archive.com/dev@.../msg01218.html
[2]: https://issues.apache.org/jira/browse/FLINK-1693

On 10 April 2015 at 12:22, Stephan Ewen <[hidden email]> wrote:

> I would remove Spargel at some point in time, but not until Gelly is out of
> staging.
>
> On Fri, Apr 10, 2015 at 12:13 PM, Ufuk Celebi <[hidden email]> wrote:
>
> > Hey all,
> >
> > currently we have two vertex-centric graph APIs: Spargel and Gelly. I
> want
> > to discuss whether we shall
> > 1) keep Spargel as a public API as it is, or
> > 2) deprecate (and remove) it.
> >
> > In my understanding, Spargel was a proof-of-concept, which stuck around.
> > It is very stable, but limited in functionality. Gelly provides a
> superset
> > of Spargel's functionality and a high-level library of graph algorithms.
> > The vertex-centric iterations actually wrap Spargel.
> >
> > I am in favour of 2):
> >
> > + Less confusing and less work to have two APIs for the same thing (we
> > have to communicate this, document it etc.)
> > + Gelly is actively maintained and getting a lot of contributions
> > - Spargel users will have to move to Gelly at some point in time (I think
> > this will happen anyways and it should be straight forward)
> >
> > The Spargel internal code will probably stick around as part of Gelly.
> The
> > question is whether we want to have two public APIs for this.
> >
> > – Ufuk
>

[Ufuk Celebi-2]

Thanks for the links, Vasia! I forgot about this.

Consensus is to deprecate and remove after Gelly is out of staging then. :-)

On Fri, Apr 10, 2015 at 12:41 PM, Vasiliki Kalavri <
[hidden email]> wrote:

> Hey!
>
> We've had this discussion before actually [1] and we also have a JIRA to
> deprecate Spargel [2] :-)
> I agree that we should keep Spargel deprecated until Gelly is out of
> staging.
>
> Cheers,
> -Vasia.
>
> [1]: https://www.mail-archive.com/dev@.../msg01218.html
> [2]: https://issues.apache.org/jira/browse/FLINK-1693
>
> On 10 April 2015 at 12:22, Stephan Ewen <[hidden email]> wrote:
>
> > I would remove Spargel at some point in time, but not until Gelly is out
> of
> > staging.
> >
> > On Fri, Apr 10, 2015 at 12:13 PM, Ufuk Celebi <[hidden email]> wrote:
> >
> > > Hey all,
> > >
> > > currently we have two vertex-centric graph APIs: Spargel and Gelly. I
> > want
> > > to discuss whether we shall
> > > 1) keep Spargel as a public API as it is, or
> > > 2) deprecate (and remove) it.
> > >
> > > In my understanding, Spargel was a proof-of-concept, which stuck
> around.
> > > It is very stable, but limited in functionality. Gelly provides a
> > superset
> > > of Spargel's functionality and a high-level library of graph
> algorithms.
> > > The vertex-centric iterations actually wrap Spargel.
> > >
> > > I am in favour of 2):
> > >
> > > + Less confusing and less work to have two APIs for the same thing (we
> > > have to communicate this, document it etc.)
> > > + Gelly is actively maintained and getting a lot of contributions
> > > - Spargel users will have to move to Gelly at some point in time (I
> think
> > > this will happen anyways and it should be straight forward)
> > >
> > > The Spargel internal code will probably stick around as part of Gelly.
> > The
> > > question is whether we want to have two public APIs for this.
> > >
> > > – Ufuk
> >
>

[Vasiliki Kalavri]

Do you think we should add a Spargel to Gelly migration guide in the docs
or something similar?

On 10 April 2015 at 13:45, Ufuk Celebi <[hidden email]> wrote:

> Thanks for the links, Vasia! I forgot about this.
>
> Consensus is to deprecate and remove after Gelly is out of staging then.
> :-)
>
> On Fri, Apr 10, 2015 at 12:41 PM, Vasiliki Kalavri <
> [hidden email]> wrote:
>
> > Hey!
> >
> > We've had this discussion before actually [1] and we also have a JIRA to
> > deprecate Spargel [2] :-)
> > I agree that we should keep Spargel deprecated until Gelly is out of
> > staging.
> >
> > Cheers,
> > -Vasia.
> >
> > [1]: https://www.mail-archive.com/dev@.../msg01218.html
> > [2]: https://issues.apache.org/jira/browse/FLINK-1693
> >
> > On 10 April 2015 at 12:22, Stephan Ewen <[hidden email]> wrote:
> >
> > > I would remove Spargel at some point in time, but not until Gelly is
> out
> > of
> > > staging.
> > >
> > > On Fri, Apr 10, 2015 at 12:13 PM, Ufuk Celebi <[hidden email]> wrote:
> > >
> > > > Hey all,
> > > >
> > > > currently we have two vertex-centric graph APIs: Spargel and Gelly. I
> > > want
> > > > to discuss whether we shall
> > > > 1) keep Spargel as a public API as it is, or
> > > > 2) deprecate (and remove) it.
> > > >
> > > > In my understanding, Spargel was a proof-of-concept, which stuck
> > around.
> > > > It is very stable, but limited in functionality. Gelly provides a
> > > superset
> > > > of Spargel's functionality and a high-level library of graph
> > algorithms.
> > > > The vertex-centric iterations actually wrap Spargel.
> > > >
> > > > I am in favour of 2):
> > > >
> > > > + Less confusing and less work to have two APIs for the same thing
> (we
> > > > have to communicate this, document it etc.)
> > > > + Gelly is actively maintained and getting a lot of contributions
> > > > - Spargel users will have to move to Gelly at some point in time (I
> > think
> > > > this will happen anyways and it should be straight forward)
> > > >
> > > > The Spargel internal code will probably stick around as part of
> Gelly.
> > > The
> > > > question is whether we want to have two public APIs for this.
> > > >
> > > > – Ufuk
> > >
> >
>

[Ufuk Celebi-2]

Yes, I do. :-) I was just going to propose that.

On Fri, Apr 10, 2015 at 2:44 PM, Vasiliki Kalavri <[hidden email]
> wrote:

> Do you think we should add a Spargel to Gelly migration guide in the docs
> or something similar?
>
> On 10 April 2015 at 13:45, Ufuk Celebi <[hidden email]> wrote:
>
> > Thanks for the links, Vasia! I forgot about this.
> >
> > Consensus is to deprecate and remove after Gelly is out of staging then.
> > :-)
> >
> > On Fri, Apr 10, 2015 at 12:41 PM, Vasiliki Kalavri <
> > [hidden email]> wrote:
> >
> > > Hey!
> > >
> > > We've had this discussion before actually [1] and we also have a JIRA
> to
> > > deprecate Spargel [2] :-)
> > > I agree that we should keep Spargel deprecated until Gelly is out of
> > > staging.
> > >
> > > Cheers,
> > > -Vasia.
> > >
> > > [1]: https://www.mail-archive.com/dev@.../msg01218.html
> > > [2]: https://issues.apache.org/jira/browse/FLINK-1693
> > >
> > > On 10 April 2015 at 12:22, Stephan Ewen <[hidden email]> wrote:
> > >
> > > > I would remove Spargel at some point in time, but not until Gelly is
> > out
> > > of
> > > > staging.
> > > >
> > > > On Fri, Apr 10, 2015 at 12:13 PM, Ufuk Celebi <[hidden email]>
> wrote:
> > > >
> > > > > Hey all,
> > > > >
> > > > > currently we have two vertex-centric graph APIs: Spargel and
> Gelly. I
> > > > want
> > > > > to discuss whether we shall
> > > > > 1) keep Spargel as a public API as it is, or
> > > > > 2) deprecate (and remove) it.
> > > > >
> > > > > In my understanding, Spargel was a proof-of-concept, which stuck
> > > around.
> > > > > It is very stable, but limited in functionality. Gelly provides a
> > > > superset
> > > > > of Spargel's functionality and a high-level library of graph
> > > algorithms.
> > > > > The vertex-centric iterations actually wrap Spargel.
> > > > >
> > > > > I am in favour of 2):
> > > > >
> > > > > + Less confusing and less work to have two APIs for the same thing
> > (we
> > > > > have to communicate this, document it etc.)
> > > > > + Gelly is actively maintained and getting a lot of contributions
> > > > > - Spargel users will have to move to Gelly at some point in time (I
> > > think
> > > > > this will happen anyways and it should be straight forward)
> > > > >
> > > > > The Spargel internal code will probably stick around as part of
> > Gelly.
> > > > The
> > > > > question is whether we want to have two public APIs for this.
> > > > >
> > > > > – Ufuk
> > > >
> > >
> >
>