Postgres vs MongoDB on modeling recursive multi-child relations - ruby-on-rails

Question

I am considering using either of the following stack for a personal project:
Nodes.js/MongoDB (learning)
Rails/Postgres (more familiar)
I would like to give MongoDB a try for learning purposes, but I am unsure if it is suitable for this problem. I would like to hear the trade-off and examples based on the following problem description, some specific questions are at the bottom:
There are a list of Products, let's say p1, p2, p3, and each product has the fields for some environmental impact, let's say A, B, C.
p1 p2
+ +
| |
| |
+------------------+ +----+----+
| | | | |
+ + + + +
p3 p4 p5 p3 p6
+ + |
| | |
+-----+-+ +---+--+ +---+--+
+ + + + + +
p7 p8 p2 p9 p10 p11
p1.A = p3.A + p4.A + p5.A
p1.B = p3.B + p4.B + p5.B
p3.A = p7.A + p8.A
Product Table would look something like this
id A B C parents children
1 4 5 6 [] [3, 4, 5]
2 10 11 12 [4] [3, 6]
3 6 7 8 [1,2] [7, 8]
4 3 9 6 [1] [2, 9]
5 3 3 10 [1] [10, 11]
6 3 1 2 [2] []
7 4 5 0 [3] []
...
Updates Process would look like this:
p1 is made of p2 and p3.
p2 is also made of p3
If p3 A, B, or C updates, it would trigger a p1 update to recalculate its A, B, C, although maybe still with old p2's value. Then when p3 updates p2, p2 updates will trigger the p1 updates again. There could be some redundant operations in the updates depending on the ordering. I am guessing that is ok.
Since the environmental impact is not a critical data, I am just looking that the data becomes eventually consistent.
In terms of scale, maybe tens of thousands of products at some point.
Questions:
1) I need to way to prevent infinite update cycle in a circular graph.
2) Can you handle this type of two-way associations in MongoDB easily, product has parents that are products, and children that are products.
3) What are the different approaches I can structure my data instead of parents and child arrays, and design this update process efficiently. If I design it such that when one product update, trigger another update, which trigger another update and the chain goes on, that could potentially make a long web request cycle?
Thanks.

Answer 1

Your model is best described as a directed graph
G = (V,E) ; V->P ; E = VxV
Therefore neither PostgreSQL or MongoDB is really good for your use case.
The biggest advantage of MongoDB compared to traditional RDBMS like PostgreSQL is the dynamic schema of MongoDB. That means you can add new records with various structure without re-defining the database schema.
But the model you have described is pretty static to me. So the argument doesn't count for your problem.
As far as I am concerned the best technology decision in your case is to use a graph database like Neo4j.
As alternative inspiration you could take a look on graph data structures. Therefore one efficient way to model a graph is the use of an adjacency matrix.