FHIR Chat · bx · mapping-framework

Stream: mapping-framework

Topic: bx


view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 16:04):

I'm doing my investigation about bidirectional transformations, as an important special case. We need it for storage & analytic transformations. Anybody interested in bx?

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 16:06):

Here are some interesting links i found:

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 16:07):

And my ideas - https://github.com/niquola/bxb/blob/master/README.md

view this post on Zulip Grahame Grieve (Jan 17 2019 at 16:15):

personally, I am skeptical about bi-directional transforms. If:

  • the value domains of both source and target are the same
  • there is no structural rework in the transformation

then bi-directional transforms are easy to specify.

view this post on Zulip Grahame Grieve (Jan 17 2019 at 16:16):

but once the complexity grows, I find that trying to specify bi-directional transforms rapidly becomes overly complicated

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 22:51):

But this is an important subset of transformations - for example, a transformation between FHIR versions can be mostly bidirectional, as well for storage & analytic or user interface i can imagine more convenient representations. Understanding the limits of bidirectional transformation can help separate information from representation.

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 22:52):

I found bx community - they even have conferences - http://bx-community.wikidot.com/start :)

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 22:54):

I like how flexible is JSON-LD representation and spec internally has kinda isomorphic transformation logic.

view this post on Zulip Grahame Grieve (Jan 17 2019 at 23:24):

a transformation between FHIR versions can be mostly bidirectional

hollow laughter on my part....

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:26):

Why not?

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:28):

I understand that in case of deep redesign of resources, they will loose this feature - but for most of resources, i think, it's possible

view this post on Zulip Grahame Grieve (Jan 17 2019 at 23:29):

I suppose it depends on what "mostly" means. The more mature resources, it's mostly true. But not always. The less mature resources, they're a dogs breakfast. (I wrote 80% of the transforms...)

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:32):

For less mature resources, i'm afraid, even one way transformation is impossible !?

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:33):

Can you provide example where for mature resource bx transformation is impossible?

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:35):

There is interesting lenses approach - where backward transformation possible if you keep the source data - pasted image

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:35):

https://www.seas.upenn.edu/~harmony/manual.pdf

view this post on Zulip Grahame Grieve (Jan 17 2019 at 23:37):

looks a bit like MDMI

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:38):

yes but more math involved ;)

view this post on Zulip nicola (RIO/SS) (Jan 17 2019 at 23:41):

But i still hope that bx will give us format flexibility - which is important for developers.


Last updated: Apr 12 2022 at 19:14 UTC