bandang0 / aamg Goto Github PK

This is just a list of enhancements to the generator we discussed lately.

The model generator should:

• allow spaces in literals, as in 'only one',
• handle the { ... }n grouping syntax, where n is a non-negative integer, meaning repeat this group n times,
• similarly, handle the * (zero or more) and + (one or more) suffixes for all groups.

Hi guys.

I thought of some checks we could do on the input supplied by users before we generate any random model with it. This would avoid unexpected results or crashes from the generator.

I suggest we check:

• that the grammar is not cyclic, i.e., that a non-terminal cannot end-up referencing itself,
• that the ASSETS file contains assets for every and all terminals of the grammar,
• that the grammar defines the modelrule,
• that the grammar is a correct grammar, i.e., that it conforms to a universal grammar for grammars, which we should therefore write.

Hi,
As discussed in #2 and in this Setuptools issue, the python3 setup.py test command for running all the project's tests will go sunset soon. Since we want to test thouroghly and easily, we should choose a new testing infrastructure.

The people from Setuptools point to the tox project. What do you say?

I know nothing on this so I'm open to all suggestions.

Hi,
So this is just a crazy idea of having a program generate astrophysical models for you.

For example, we know now that short gamma-ray bursts are produced in the merging of compact objects. However, this wasn't suggested till the 1980's. So imagine someone in the 1970's had a program which randomly output the string "What if short gamma-ray bursts were due to the merging of compact objects?". I reckon just reading that line and looking at the data with a new eye would have saved gamma-ray burst science a couple years.

In this project, we would simply have a list of phenomena (gamma-ray bursts, fast radio bursts, solar flares, etc.), a list of progenitor objects (neutrinos, compact objects, galaxies, clouds, etc.), a list of actions (merging, interacting, decaying, etc.), a list on environments (galaxies, globular clusters, planet atmospheres, etc.). Then the program would generate random models. Like:

$ ./aamg
$ What if fast radio bursts were due to the collision of asteroids with the magnetopheres of
pulsars?

(Actually that's not a bad idea.)

Also, we could ask to generate models of a given phenomena:

$ ./aamg --phenomena "gamma-ray bursts"
$ What if gamma-ray bursts were due to the collapse of massive stars in spiral galaxies?
$ ./aamg --phenomena "gamma-ray bursts"
$ What if gamma-ray bursts were due to explosions of exoplanets in close orbit with white dwarfs?

Also, we could vary the sentence structure and ask for specific progenitors:

$ ./aamg --progenitor "M-type star"
$ M-type stars produce neutrinos from their envelops during Helium shell burning.
$ ./aamg --progenitor "M-type star"
$ Active galactic nuclei flares are caused by merging of M-type stars with black holes.

I feel like there are some language issues because we want to have correct English outputs every time. Also some actions (like merging) require two progenitors (like a star and a black hole), whereas others (like collapsing) require only one. Certainly there is some rigid structure (like subject-verb-complement) we have to pin down so that everything comes out smoothly!

On the long run, I think we should do the core programming with some elementary bits and then let people populate the possible models by adding objects and actions etc. through PR.

Tell me what you think about the implementation of this! 💫

Hi,
So before we enhance the project I suggest that we kick-start the documentation real quick, that is:

• Start the CHANGELOG at the end of README.md
• Write a very small description and first usage in README.md
• Go through the project and update the present (module and function) docstrings just to make it all a little cleaner.