You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@ossie.apache.org by Yufei Gu <fl...@gmail.com> on 2026/08/25 01:11:39 UTC
[DISCUSS] Consolidating the Python converters
Hi all,
I’ve been looking at how the converters are organized today. Each Python
converter is a separate project with its own pyproject.toml, Python
version, dependencies, CLI, and test setup. This makes the repository feel
more like a collection of independent projects than a single Ossie
toolchain.
Would it make sense to consolidate the Python converters into a single
project? We could keep each converter as its own module while sharing:
1. One build and test setup
2. A common converter interface and issue format
3. A consistent CLI, for example:
ossie-convert import databricks -i view.yaml -o model.yaml
ossie-convert export snowflake -i model.yaml -o model.snowflake.yaml
Heavier dependencies, such as MetricFlow or sqlglot, could remain optional
extras. Existing commands could stay as compatibility wrappers during the
transition.
I’m not suggesting rewriting the Java converters. We could take a similar
approach there: one Java project with separate modules.
The main tradeoff is that the Python converters would share a release
cadence. Longer term, more complex vendor-maintained converters may be
better placed in separate repositories rather than the main Ossie
repository.
Curious to hear what others think.
Yufei
Re: [DISCUSS] Consolidating the Python converters
Posted by Yufei Gu <fl...@gmail.com>.
Thanks JB for the additional context.
Other than the release and review concern, I think there are common
components worth sharing. For example, the logic for producing a specific
version of the Ossie semantic model could be reused across multiple
converters. This shared layer would become even more valuable as different
Ossie versions coexist and need to be supported.
That said, creating a separate repository for each converter, as JB
suggested offline, also sounds like a good way forward. It would keep the
core repository focused while still allowing each converter
evolve separately.
Yufei
On Mon, Aug 24, 2026 at 11:01 PM Jean-Baptiste Onofré <jb...@nanthrax.net>
wrote:
> Hi Yufei,
>
> This topic was previously discussed with Yong and others.
>
> To encourage contributions, we intentionally decided to allow maximum
> autonomy for each converter, giving them their own approach and
> release cycle. The goal is to maintain a "collection" of independent
> converters rather than a unified toolchain. Regarding the release
> process, we agreed to have a single, independent release per
> converter.
>
> While converters can share a common layer, this should not be
> enforced, and we must maintain the ability to release each converter
> independently.
>
> Regards,
> JB
>
> On Tue, Aug 25, 2026 at 3:12 AM Yufei Gu <fl...@gmail.com> wrote:
> >
> > Hi all,
> >
> > I’ve been looking at how the converters are organized today. Each Python
> > converter is a separate project with its own pyproject.toml, Python
> > version, dependencies, CLI, and test setup. This makes the repository
> feel
> > more like a collection of independent projects than a single Ossie
> > toolchain.
> >
> > Would it make sense to consolidate the Python converters into a single
> > project? We could keep each converter as its own module while sharing:
> >
> > 1. One build and test setup
> > 2. A common converter interface and issue format
> > 3. A consistent CLI, for example:
> >
> > ossie-convert import databricks -i view.yaml -o model.yaml
> > ossie-convert export snowflake -i model.yaml -o model.snowflake.yaml
> >
> > Heavier dependencies, such as MetricFlow or sqlglot, could remain
> optional
> > extras. Existing commands could stay as compatibility wrappers during the
> > transition.
> >
> > I’m not suggesting rewriting the Java converters. We could take a similar
> > approach there: one Java project with separate modules.
> >
> > The main tradeoff is that the Python converters would share a release
> > cadence. Longer term, more complex vendor-maintained converters may be
> > better placed in separate repositories rather than the main Ossie
> > repository.
> >
> > Curious to hear what others think.
> >
> > Yufei
>
Re: [DISCUSS] Consolidating the Python converters
Posted by Jean-Baptiste Onofré <jb...@nanthrax.net>.
Hi Yufei,
This topic was previously discussed with Yong and others.
To encourage contributions, we intentionally decided to allow maximum
autonomy for each converter, giving them their own approach and
release cycle. The goal is to maintain a "collection" of independent
converters rather than a unified toolchain. Regarding the release
process, we agreed to have a single, independent release per
converter.
While converters can share a common layer, this should not be
enforced, and we must maintain the ability to release each converter
independently.
Regards,
JB
On Tue, Aug 25, 2026 at 3:12 AM Yufei Gu <fl...@gmail.com> wrote:
>
> Hi all,
>
> I’ve been looking at how the converters are organized today. Each Python
> converter is a separate project with its own pyproject.toml, Python
> version, dependencies, CLI, and test setup. This makes the repository feel
> more like a collection of independent projects than a single Ossie
> toolchain.
>
> Would it make sense to consolidate the Python converters into a single
> project? We could keep each converter as its own module while sharing:
>
> 1. One build and test setup
> 2. A common converter interface and issue format
> 3. A consistent CLI, for example:
>
> ossie-convert import databricks -i view.yaml -o model.yaml
> ossie-convert export snowflake -i model.yaml -o model.snowflake.yaml
>
> Heavier dependencies, such as MetricFlow or sqlglot, could remain optional
> extras. Existing commands could stay as compatibility wrappers during the
> transition.
>
> I’m not suggesting rewriting the Java converters. We could take a similar
> approach there: one Java project with separate modules.
>
> The main tradeoff is that the Python converters would share a release
> cadence. Longer term, more complex vendor-maintained converters may be
> better placed in separate repositories rather than the main Ossie
> repository.
>
> Curious to hear what others think.
>
> Yufei