Hi Josh,nice to see you drop in. Sending the diff back would be amazing and
something I always wanted as well. I suppose we'd need to figure out what
conflict resolution looked like for something like that. I would think that
to be the hard part.

On Thu, Jun 4, 2026 at 6:15 PM Josh Perryman <[email protected]> wrote:

> I, for one, love this idea.
>
> I've had several use cases in the past where there was a large "global"
> graph, but then a "local" (personalized, sliced, whatever) graph was
> extracted for in-memory analysis.
>
> For a 2.0 I would love the ability to apply the "diff" of a mutated graph
> back to the global graph.
>
> -Josh
>
>
>
> On Thu, Jun 4, 2026 at 3:40 PM Stephen Mallette <[email protected]>
> wrote:
>
> > Recent work on gremlin-mcp which now has access to the ANTLR grammar and
> > the introduction of proper subgraph() support in GLVs, particularly
> > javascript, had me thinking about an old idea of codifying a subset of
> > Gremlin that would be implemented as a processing engine outside of
> > Javascript. The primary use case would be to use subgraph() to return a
> > slice of a remote graph and then use the subset of Gremlin to traverse
> that
> > subgraph locally on the client side. I think that core use case kicks
> open
> > a lot of doors for applications.. Secondly, I think I've seen a number of
> > Gremlin "lite" or "inspired" graph query syntaxes out there over the
> years
> > and perhaps standardizing around what that idea means could open up wider
> > opportunities for those systems to be more aligned with the official
> > TinkerPop community.
> >
> > i sense these initial steps would probably be:
> >
> > Source
> > V, E
> >
> > Navigation
> > out, in, both, outE, inE, bothE, outV, inV, otherV
> >
> > Filter
> > has, hasId, hasLabel, hasNot, hasKey, hasValue
> >
> > Value Extraction
> > id, label, values, valueMap, elementMap, value, key
> >
> > Path
> > path
> >
> > Range
> > limit, range, skip, tail
> >
> > Ordering
> > order
> >
> > Mutation
> > addV, addE (with from/to), property
> >
> > Any additional processing you might do would have to fall back to lambda
> > style processing like via forEach or reduce.
> >
> > Anyway, I think we could call this diminutive standard Tiny Gremlin and
> > perhaps just start with Javascript as some kind of proof of concept.
> >
>

Reply via email to