Transforming objects

Dear Pharo users,

I develop software dedicated to the simulation of plant development and
3D architecture.

In the past, I used Java, but I am very interested in translating code
in Pharo. I appreciated very much the language and the course.

I have a question for you all :

During plant development (ontogeny), new organs appear, that are
gradually tranformed. For example, a bud can make a flower, and then a
fruit.

Organs are conveniently modelled as classes. During the course, I
learned that it was better to make several classes (bud, flower, fruit)
instead of a "fat class" such as "reproductive organ" with many
attributes and stages (bud stage, flower stage, fruit stage). I wonder
how to transform an object into another one, to represent for example
the fact that the flower may become a fruit.

Maybe a naive question ?

Thanks for your lights !

Loïc

Loïc,

Am I correct in assuming you are thinking about a model in which a plant
can have a growing and shrinking list of organs over time and some of
them even have (a growing list of) stages?

I am not an expert on the matter, but to me the stages idea sounds good.
It even allows for historic information (it was a bud from Date1 to
Date2 and since then it is a flower). So transformation would be adding
a new Stage and "terminate" the previous one.

With this you can ask a plant about its current stage on the
reproductive front and maybe other organs as well at any point time.
This may be helpful if you want to build an animation of the plant's
development over time. You can have separate life cycles for each organ.

Does the plant transform in the real world? I mean, is a cherry tree
still a cherry tree, even if it's not blooming any more and carries
fruit instead? So why transform the object in its abstract model?


Just a thought, maybe better ideas show up.

Joachim


Am 16.01.19 um 10:15 schrieb Loïc Pagès:

Hi Joachim,

Thank you for your reply and for your ideas. I will reply to your
questions below.

Le 16/01/2019 à 10:25, [hidden email] a écrit :
> Loïc,
>
> Am I correct in assuming you are thinking about a model in which a
> plant can have a growing and shrinking list of organs over time and
> some of them even have (a growing list of) stages?

Yes, the plant can be seen as a growing and shrinking list of organs.
Very short list at the beginning, and becoming larger. Some organs can
fall (leaves, roots) or even be pruned (shoots in fruit trees), so that
shrinking can also occur.

However, the list of stages for each organ is usually known at the
beginning, for each type of organ. It depends of knowledge and level of
detail that we include in the model.

>
> I am not an expert on the matter, but to me the stages idea sounds
> good. It even allows for historic information (it was a bud from Date1
> to Date2 and since then it is a flower). So transformation would be
> adding a new Stage and "terminate" the previous one.

This can be an approach. But the conditions for passing from one stage
to another can be more complex and involve several aspects (age,
previous growth, pollination, etc) with a lot of code. So, if you keep
this code into the general "reproductive organ" class, you make a "fat
class" with a lot of tests. This was not recommended in the Pharo course
:-)

>
> With this you can ask a plant about its current stage on the
> reproductive front and maybe other organs as well at any point time.
> This may be helpful if you want to build an animation of the plant's
> development over time. You can have separate life cycles for each organ.
This is actually an objective of such models, to be able to get the
status of the whole plant, regarding the stages of organs.
>
> Does the plant transform in the real world? I mean, is a cherry tree
> still a cherry tree, even if it's not blooming any more and carries
> fruit instead? So why transform the object in its abstract model?
The plant usually changes because of its development, but a cherry tree
remains a cherry tree !  And flowers are necessary steps to get fruits
!  The idea was not to transform the whole plant, but each individual
organ that it carries.
>
>
> Just a thought, maybe better ideas show up.

OK, thank you again

Loïc

It seems to me that in the real world, a bud does not get replaced by a flower

nor a flower by a fruit, nor is there a definite point at which a thing stops

being a bud and starts being a flower, and so on.  It's summer now, but spring

wasn't that long ago, and I enjoyed watching the imperceptible *continuous*

changes.  So anything you do is going to be imposing discontinuity on a more-

or-less continuous process and is definitely going to be an approximation.

You may perhaps be looking for the "State" design pattern where there is an

object with a fixed identity (*this* bud/flower/fruit) which behaves

differently at different times, and this is modelled by the "outer" object

delegating some of its behaviour to a "state" which can be replaced.

So you would have

  PlantOrgan

    PlantReproductiveOrgan (state)

  PlantReproductiveOrganState

    Bud

    Flower

    DeadFlower

    ImmatureFruit

    MatureFruit

You will find abundant material about the "State" design pattern on the web.

On Thu, 17 Jan 2019 at 00:29, Loïc Pagès <[hidden email]> wrote:

Hi Joachim,

Thank you for your reply and for your ideas. I will reply to your
questions below.

Le 16/01/2019 à 10:25, [hidden email] a écrit :
> Loïc,
>
> Am I correct in assuming you are thinking about a model in which a
> plant can have a growing and shrinking list of organs over time and
> some of them even have (a growing list of) stages?

Yes, the plant can be seen as a growing and shrinking list of organs.
Very short list at the beginning, and becoming larger. Some organs can
fall (leaves, roots) or even be pruned (shoots in fruit trees), so that
shrinking can also occur.

However, the list of stages for each organ is usually known at the
beginning, for each type of organ. It depends of knowledge and level of
detail that we include in the model.

>
> I am not an expert on the matter, but to me the stages idea sounds
> good. It even allows for historic information (it was a bud from Date1
> to Date2 and since then it is a flower). So transformation would be
> adding a new Stage and "terminate" the previous one.

This can be an approach. But the conditions for passing from one stage
to another can be more complex and involve several aspects (age,
previous growth, pollination, etc) with a lot of code. So, if you keep
this code into the general "reproductive organ" class, you make a "fat
class" with a lot of tests. This was not recommended in the Pharo course
:-)

>
> With this you can ask a plant about its current stage on the
> reproductive front and maybe other organs as well at any point time.
> This may be helpful if you want to build an animation of the plant's
> development over time. You can have separate life cycles for each organ.
This is actually an objective of such models, to be able to get the
status of the whole plant, regarding the stages of organs.
>
> Does the plant transform in the real world? I mean, is a cherry tree
> still a cherry tree, even if it's not blooming any more and carries
> fruit instead? So why transform the object in its abstract model?
The plant usually changes because of its development, but a cherry tree
remains a cherry tree !  And flowers are necessary steps to get fruits
!  The idea was not to transform the whole plant, but each individual
organ that it carries.
>
>
> Just a thought, maybe better ideas show up.

OK, thank you again

Loïc

>
>
>
> Am 16.01.19 um 10:15 schrieb Loïc Pagès:
>> Dear Pharo users,
>>
>> I develop software dedicated to the simulation of plant development
>> and 3D architecture.
>>
>> In the past, I used Java, but I am very interested in translating
>> code in Pharo. I appreciated very much the language and the course.
>>
>> I have a question for you all :
>>
>> During plant development (ontogeny), new organs appear, that are
>> gradually tranformed. For example, a bud can make a flower, and then
>> a fruit.
>>
>> Organs are conveniently modelled as classes. During the course, I
>> learned that it was better to make several classes (bud, flower,
>> fruit) instead of a "fat class" such as "reproductive organ" with
>> many attributes and stages (bud stage, flower stage, fruit stage). I
>> wonder how to transform an object into another one, to represent for
>> example the fact that the flower may become a fruit.
>>
>> Maybe a naive question ?
>>
>> Thanks for your lights !
>>
>> Loïc
>>
>>
>>
>>
>
>

El mié., 16 ene. 2019 a las 14:49, Ben Coman via Pharo-users (<[hidden email]>) escribió:

On Wed, 16 Jan 2019 at 17:16, Loïc Pagès <[hidden email]> wrote:

Maybe a naive question ?

Maybe #become: ...

https://gbracha.blogspot.com/2009/07/miracle-of-become.html

Although miraculous, using #become: to model a business domain is harmful, and has more undesired side effects than benefits.

So I would save it to a very limited use case scenarios, which are more related with low level stuff.

Esteban A. Maringolo  

 

This is a challenging modelling issue, life progress or any other dissipative/emergent structure are "restricted" by the boundaries of your objects (e.g. if you model a flock, and it divides into two flocks, and then merges again, which flock remains? Note: I wouldn't model the Flock at all).

So it is hard to advice without knowing exactly what the use case is or how the simulation is done,  but in the past I had to model a non-organic tree structure that was changing on a daily basis, where each part had it's own attributes and we should be able to query (aka "reconstruct") the "tree structure" at any point in time and query specific parts history, my solution to that was to have an "invariant" identity (E.g. "Plant #1234") and a subtree of elements, where each child wasn't the "organ" directly, but a "node" that contains the "organ" with a starting and an end date (like Joachim suggested) and know the parent "organ" (or the plant itself). In your case, the organs are not going to be transplanted (as was in my case), so the organ/node could be merged into one, which simplifies everything.

I don't know if your model has a tree like structure, whether it's a lemon tree or just a dandelion :), so I don't know if, e.g., the flower contains the fruit, or as I understand biology, the plant contains the flower until some point where it stops being a flower and then you replace it by a fruit.

I hope it helps you.

Esteban A. Maringolo

El mié., 16 ene. 2019 a las 9:19, Richard O'Keefe (<[hidden email]>) escribió:

It seems to me that in the real world, a bud does not get replaced by a flower

nor a flower by a fruit, nor is there a definite point at which a thing stops

being a bud and starts being a flower, and so on.  It's summer now, but spring

wasn't that long ago, and I enjoyed watching the imperceptible *continuous*

changes.  So anything you do is going to be imposing discontinuity on a more-

or-less continuous process and is definitely going to be an approximation.

You may perhaps be looking for the "State" design pattern where there is an

object with a fixed identity (*this* bud/flower/fruit) which behaves

differently at different times, and this is modelled by the "outer" object

delegating some of its behaviour to a "state" which can be replaced.

So you would have

  PlantOrgan

    PlantReproductiveOrgan (state)

  PlantReproductiveOrganState

    Bud

    Flower

    DeadFlower

    ImmatureFruit

    MatureFruit

You will find abundant material about the "State" design pattern on the web.

On Thu, 17 Jan 2019 at 00:29, Loïc Pagès <[hidden email]> wrote:

Hi Joachim,

Thank you for your reply and for your ideas. I will reply to your
questions below.

Le 16/01/2019 à 10:25, [hidden email] a écrit :
> Loïc,
>
> Am I correct in assuming you are thinking about a model in which a
> plant can have a growing and shrinking list of organs over time and
> some of them even have (a growing list of) stages?

Yes, the plant can be seen as a growing and shrinking list of organs.
Very short list at the beginning, and becoming larger. Some organs can
fall (leaves, roots) or even be pruned (shoots in fruit trees), so that
shrinking can also occur.

However, the list of stages for each organ is usually known at the
beginning, for each type of organ. It depends of knowledge and level of
detail that we include in the model.

>
> I am not an expert on the matter, but to me the stages idea sounds
> good. It even allows for historic information (it was a bud from Date1
> to Date2 and since then it is a flower). So transformation would be
> adding a new Stage and "terminate" the previous one.

This can be an approach. But the conditions for passing from one stage
to another can be more complex and involve several aspects (age,
previous growth, pollination, etc) with a lot of code. So, if you keep
this code into the general "reproductive organ" class, you make a "fat
class" with a lot of tests. This was not recommended in the Pharo course
:-)

>
> With this you can ask a plant about its current stage on the
> reproductive front and maybe other organs as well at any point time.
> This may be helpful if you want to build an animation of the plant's
> development over time. You can have separate life cycles for each organ.
This is actually an objective of such models, to be able to get the
status of the whole plant, regarding the stages of organs.
>
> Does the plant transform in the real world? I mean, is a cherry tree
> still a cherry tree, even if it's not blooming any more and carries
> fruit instead? So why transform the object in its abstract model?
The plant usually changes because of its development, but a cherry tree
remains a cherry tree !  And flowers are necessary steps to get fruits
!  The idea was not to transform the whole plant, but each individual
organ that it carries.
>
>
> Just a thought, maybe better ideas show up.

OK, thank you again

Loïc

>
>
>
> Am 16.01.19 um 10:15 schrieb Loïc Pagès:
>> Dear Pharo users,
>>
>> I develop software dedicated to the simulation of plant development
>> and 3D architecture.
>>
>> In the past, I used Java, but I am very interested in translating
>> code in Pharo. I appreciated very much the language and the course.
>>
>> I have a question for you all :
>>
>> During plant development (ontogeny), new organs appear, that are
>> gradually tranformed. For example, a bud can make a flower, and then
>> a fruit.
>>
>> Organs are conveniently modelled as classes. During the course, I
>> learned that it was better to make several classes (bud, flower,
>> fruit) instead of a "fat class" such as "reproductive organ" with
>> many attributes and stages (bud stage, flower stage, fruit stage). I
>> wonder how to transform an object into another one, to represent for
>> example the fact that the flower may become a fruit.
>>
>> Maybe a naive question ?
>>
>> Thanks for your lights !
>>
>> Loïc
>>
>>
>>
>>
>
>

Hi,

I would use an ad-hoc solution, modeling the Plant tree structure as a
document tree in Grafoscopio [1]. The root of the plant representation
would be the root of the document, the branches and sub-branches would
be nodes and sub-nodes in the documents, respectively. And each node
would contain a dictionary with the properties I want to reflect for
that particular node. The header of the node would be a shortcut to the
more detailed representation in the node body (the dictionary). Because
Grafoscopio notebooks are just plain STON text, you could store plants
history on a DVCS (Fossil, Git) and reify that ad-hoc model with methods
that capture what started just as dictionaries.

[1] http://mutabit.com/grafoscopio/index.en.html

Notice that Grafoscopio documents are programmable, so, while
Grafoscopio provides a GUI for editing them, you could access the API
and create them by scripting/coding.

Also the long, painful, joyful Karma of my PhD  is finishing with only
public dissertation pending, so I will have more time to care for
Grafoscopio source code and issues (hopefully I will send some updates,
once the deserved quiet time after intense thesis writing have finished).

Cheers,

Offray

On 16/1/19 4:15, Loïc Pagès wrote: