composition: Add section for up-converting unflattened models #41
Conversation
EricCousineau-TRI
force-pushed
the
feature-sdformat-pr355-up-convert
branch
from
912ffce
to
b0d39d4
Compare
There was a problem hiding this comment.
Reviewed 1 of 1 files at r1.
Reviewable status: 0 of 1 files reviewed, 7 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 61 at r1 (raw file):
behavior than direct nesting with `//model/model`, part of the reason being that nesting directly with `//model/model` was not implemented in the `libsdformat` until 9.3.0. As mentioned in the
nit: in the libsdformat -> in libsdformat
composition/proposal.md, line 64 at r1 (raw file):
[Legacy Behavior](/tutorials?tut=composition&ver=1.5&#libsdformats-implementation-of-include-in-models) documentation, `//include` works by effectively flattening the model; when this happens, certain details may "leak" through. Normall, nesting of models generally implies that the elements can be
Normally
composition/proposal.md, line 583 at r1 (raw file):
`M1::my_link`) should be "unnested" (e.g. `my_link` in model `M1`). To illustrate, the following model from the Legacy Behavior documentation:
add hyperlink? http://sdformat.org/tutorials?tut=composition&cat=specification&#libsdformats-implementation-of-include-in-models
composition/proposal.md, line 597 at r1 (raw file):
</visual> </link> <link name="ChildModel::L2"/>
I think the example in the legacy behavior documentation may be wrong; I think this link would have a pose of 1 0 1 0 0 0
composition/proposal.md, line 662 at r1 (raw file):
* Since the naively converted file is neither SDFormat 1.5 nor 1.8, it is indicated as SDFormat `1.5*`, meaning that it's an intermediate format which namely adheres to SDFormat 1.5, but has directly nested models.
SDFormat 1.5 supports directly nested models in the spec, so I think we could leave it as 1.5
composition/proposal.md, line 665 at r1 (raw file):
* Since `libsdformat9.3` supported direct nesting but `//include` still worked via flattening, unflattening will occur regardless of whether or not there are directly nested models in the model, and the unflattening my handle
my -> may
composition/proposal.md, line 670 at r1 (raw file):
the newly created `//model/pose`. * This offers simplicity and ensures that the effective `__model__` frame will coincide with the newly created nested models' `__model__` frames.
as of libsdformat 9.2, the a flattened model would also have an injected ChildModel::__model__ per bitbucket pr 668; perhaps we need to upgrade the legacy behavior documentation a bit; this reasoning is valid for versions < 9.2 but doesn't cover the injected __model__ frames
There was a problem hiding this comment.
Reviewable status: 0 of 2 files reviewed, 6 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 61 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
nit:
in the libsdformat->in libsdformat
Done.
composition/proposal.md, line 64 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
Normally
Done.
composition/proposal.md, line 583 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
add hyperlink? http://sdformat.org/tutorials?tut=composition&cat=specification&#libsdformats-implementation-of-include-in-models
Done.
composition/proposal.md, line 597 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
I think the example in the legacy behavior documentation may be wrong; I think this link would have a pose of
1 0 1 0 0 0
Done. Good catch!
composition/proposal.md, line 662 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
SDFormat 1.5 supports directly nested models in the spec, so I think we could leave it as
1.5
Done. I actually changed it to 1.4; if we think we won't run into this in the wild, then I could strip it out entirely.
composition/proposal.md, line 665 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
my->may
Done.
There was a problem hiding this comment.
Reviewable status: 0 of 2 files reviewed, 7 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 680 at r3 (raw file):
will coincide with the newly created nested models' `__model__` frames. **Example of a failing conversions**
Working: Should also try to explicitly state where this happens in high or low level parsing stages.
- readDoc - may handle the conversion, then pass off to readXml
- readXml - should handle the latest version
There was a problem hiding this comment.
Reviewable status: 0 of 2 files reviewed, 7 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 670 at r1 (raw file):
Previously, scpeters (Steve Peters) wrote…
as of
libsdformat 9.2, the a flattened model would also have an injectedChildModel::__model__per bitbucket pr 668; perhaps we need to upgrade the legacy behavior documentation a bit; this reasoning is valid for versions < 9.2 but doesn't cover the injected__model__frames
Working: Will update examples (legacy and here) that incorporate the __model__ frames.
There was a problem hiding this comment.
Reviewed 2 of 2 files at r3.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 670 at r1 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Working: Will update examples (legacy and here) that incorporate the
__model__frames.
Will first confirm what happens with SDFormat 1.4 and include.
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 670 at r1 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Will first confirm what happens with
SDFormat 1.4and include.
Will modified tests introduced here:
https://github.com/osrf/sdformat/compare/cc1f9b00e31b991a71343db5afbe3665d8af1ff8..4836e3d2577fe4e7fe52a94805bc8a9ba6bb432d
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 670 at r1 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Will modified tests introduced here:
https://github.com/osrf/sdformat/compare/cc1f9b00e31b991a71343db5afbe3665d8af1ff8..4836e3d2577fe4e7fe52a94805bc8a9ba6bb432d
Setup:
Using this workspace setup: EricCousineau-TRI/repro@0162a3d
colcon build
cd build/sdformat11
make INTEGRATION_nested_model # For intermediate changes.
./test/integration/INTEGRATION_nested_model --gtest_filter=NestedModel.NestedModelWithFramesDirectComparison
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 670 at r1 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Setup:
Using this workspace setup: EricCousineau-TRI/repro@0162a3d
colcon build cd build/sdformat11 make INTEGRATION_nested_model # For intermediate changes. ./test/integration/INTEGRATION_nested_model --gtest_filter=NestedModel.NestedModelWithFramesDirectComparison
With this branch: gazebosim/sdformat@master...EricCousineau-TRI:feature-composition-model-inject-wip | gazebosim/sdformat@3f46da9
Some minor notes (related to above) discussion - but setting /sdf/@version="1.4" doesn't seem to trip anything up in the parsing setup.
Makes senese, just wanted to confirm.
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 670 at r1 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
With this branch: gazebosim/sdformat@master...EricCousineau-TRI:feature-composition-model-inject-wip | gazebosim/sdformat@3f46da9
Some minor notes (related to above) discussion - but setting
/sdf/@version="1.4"doesn't seem to trip anything up in the parsing setup.
Makes senese, just wanted to confirm.
Here's what the flattened model looks like with SDFormat 1.8 (assuming it'll be the same for SDFormat 1.7):
There was a problem hiding this comment.
Reviewable status: 0 of 2 files reviewed, 2 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/legacy_behavior.md, line 223 at r4 (raw file):
```xml <model name="ParentModel">
BTW I've replaced this with the output of current libsdformat, but with some excess default values trimmed.
composition/proposal.md, line 670 at r1 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Here's what the flattened model looks like with SDFormat 1.8 (assuming it'll be the same for SDFormat 1.7):
Done.
composition/proposal.md, line 668 at r3 (raw file):
been emitted by `libsdformat10` by using *valid* `//include` statements. It is possible to write valid models in SDFormat 1.7 that are invalid in SDFormat 1.8; no effort will be made to reconcile those models (see examples below). * Since the naively converted file is neither SDFormat 1.5 nor 1.8, it is
BTW I've removed this section for now. Per mention below, the whole "is the version tag strictly true" seems like a separate concern.
There was a problem hiding this comment.
Reviewed 2 of 2 files at r4.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/legacy_behavior.md, line 223 at r4 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
BTW I've replaced this with the output of current
libsdformat, but with some excess default values trimmed.
looks good. How do you feel about adding a duplicate versioned 1.7 file with these changes and leaving the old version in a separate file? I can do the grunt work
composition/proposal.md, line 679 at r4 (raw file):
`ChildModel::__model__`) will have this frame removed, and this will be converted to the appropriate `//model/pose`. If `@attached_to` is specified to something other than this nexted model's first link, then
nexted -> nested
There was a problem hiding this comment.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 680 at r3 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Working: Should also try to explicitly state where this happens in high or low level parsing stages.
- readDoc - may handle the conversion, then pass off to readXml
- readXml - should handle the latest version
Working: What if we create a new conversion keyword "unnest" and do it in the Convert class? This would only be done in a 1.7->1.8 conversion.
composition/proposal.md, line 670 at r4 (raw file):
the naive up-conversion process. * This up-conversion is *only* intended to support models that *could* have been emitted by `libsdformat10` by using *valid* `//include` statements. It is possible to write valid models in SDFormat 1.7 that are invalid in SDFormat
nit: libsdformat10 -> libsdformat9.3?
There was a problem hiding this comment.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 680 at r3 (raw file):
Previously, azeey (Addisu Z. Taddese) wrote…
Working: What if we create a new conversion keyword "unnest" and do it in the
Convertclass? This would only be done in a1.7->1.8conversion.
that makes sense to me
composition/proposal.md, line 670 at r4 (raw file):
Previously, azeey (Addisu Z. Taddese) wrote…
nit:
libsdformat10->libsdformat9.3?
flattening has been happening since before libsdformat4 and its behavior changed in libsdformat 9.2 with the inclusion of bitbucket PR 668. maybe we could just omit the version and say libsdformat?
There was a problem hiding this comment.
Reviewable status: all files reviewed, 3 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 670 at r4 (raw file):
Previously, scpeters (Steve Peters) wrote…
flattening has been happening since before libsdformat4 and its behavior changed in libsdformat 9.2 with the inclusion of bitbucket PR 668. maybe we could just omit the version and say
libsdformat?
Actually, this bit of history seems useful. Will incorporate it here.
There was a problem hiding this comment.
Reviewable status: 1 of 2 files reviewed, 2 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 680 at r3 (raw file):
Previously, scpeters (Steve Peters) wrote…
that makes sense to me
Er, not sure what "conversion keyword" means. Can you point to that context?
composition/proposal.md, line 670 at r4 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Actually, this bit of history seems useful. Will incorporate it here.
Done. Incorporated libsdformat4.
However, I don't think the inclusion of pose frame semantics matters in this exact bullet point. Yes, the below model does use pose frame semantics, but an invalid model could also arise from an invalid specification of other interface elements, e.g. //joint/child, //join/parent, etc.
I've updated the example below to use pre-pose-frame semantics stuff as a better illustration.
composition/proposal.md, line 679 at r4 (raw file):
Previously, scpeters (Steve Peters) wrote…
nexted -> nested
Done. Oops!
There was a problem hiding this comment.
Reviewable status: 1 of 2 files reviewed, 2 unresolved discussions (waiting on @azeey, @EricCousineau-TRI, and @scpeters)
composition/proposal.md, line 680 at r3 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Er, not sure what "conversion keyword" means. Can you point to that context?
Yeah, also not sure what the gain is there? Can talk more f2f.
|
Per f2f:
e.g. code |
There was a problem hiding this comment.
Reviewed 1 of 1 files at r6.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 680 at r3 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Yeah, also not sure what the gain is there? Can talk more f2f.
I think we agreed on this during f2f today. Does this idea require any text changes here?
composition/proposal.md, line 83 at r6 (raw file):
legacy format. Generally, the interface between models only really needs access to explicit and implicit frames (for welding joints, attaching sensors, etc.). The present implementation of `//include` (present
nit: using "present" twice so closely sounds funny.
- "The current implementation"
or - remove "present" from the parenthetical "(since `libsdformat4)"
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/proposal.md, line 680 at r3 (raw file):
Previously, scpeters (Steve Peters) wrote…
I think we agreed on this during f2f today. Does this idea require any text changes here?
Yeah, will add a brief blurb above about the potential implementation.
There was a problem hiding this comment.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @azeey and @EricCousineau-TRI)
composition/legacy_behavior.md, line 223 at r4 (raw file):
Previously, scpeters (Steve Peters) wrote…
looks good. How do you feel about adding a duplicate versioned 1.7 file with these changes and leaving the old version in a separate file? I can do the grunt work
OK Sounds good to me! I'm assuming that'll happen in a separate PR, so will get this 'un merged in.
There was a problem hiding this comment.
Reviewable status: 1 of 2 files reviewed, 1 unresolved discussion (waiting on @azeey and @scpeters)
composition/proposal.md, line 680 at r3 (raw file):
Previously, EricCousineau-TRI (Eric Cousineau) wrote…
Yeah, will add a brief blurb above about the potential implementation.
Done. Added in Addisu's suggestion; we can change later if it becomes different.
composition/proposal.md, line 83 at r6 (raw file):
Previously, scpeters (Steve Peters) wrote…
nit: using "present" twice so closely sounds funny.
- "The current implementation"
or- remove "present" from the parenthetical "(since `libsdformat4)"
Done.
Copied legacy_behavior.md from 66d3d88 (before #41) to legacy_behavior_1-5.md Signed-off-by: Steve Peters <[email protected]>
Copied legacy_behavior.md from 66d3d88 (before #41) to legacy_behavior_1-5.md Signed-off-by: Steve Peters <[email protected]>
As motivated here: gazebosim/sdformat#355 (review)
Towards gazebosim/sdformat#342
Resolves #39
Preview:
http://sdformat.org/tutorials?tut=composition_proposal&branch=feature-sdformat-pr355-up-convert&repo=https%3A%2F%2Fgithub.com%2FEricCousineau-TRI%2Fsdf_tutorials-1#1-4-7-up-converting-flattened-models
This change is