gaiana / nirdust Goto Github PK

add to metadata the spectral region information in the case of using the default values or in the case the user changes it with 'compute_noise'. Once the region is in the metadata, change "line_spectrum" to use that info.

• copy ez_setup.py
• Copy and modify the setup.py from grispy

that

change the first output of "line_spectrum" to return a NirdustSpectrum

Study the best way to handle different spectral_axis lenght

One idea is convert cut_edges into a function that takes both spectrum and cut them in the same spectral range and then cut the extra pixels that could be left.

Other idea is to leave this issue to be solved in sp_correction but using wavelenght to compare the spectra instead of pixels (actually both have to be used)

Los archivos MANIFEST.in controlan el contenido de los paquetes binarios en python.

Por ejemplo controlan que no se agrgegue toda la documentacion, pero si la licencia al paquete (importante por cuestiones legales)

Agregar el environment de test siguiente a tox (y travis)

[testenv:check-manifest]
skip_install = True
usedevelop = False
deps = check-manifest
commands =
    check-manifest

Puede crearse un archivo MANIFEST vacio con check-manifest -c es importante tener en cuenta que para que todo funcione tienen que hacer git add MANIFEST.in

Ver MANIFEST.in de pyedra

La clase NirdustSpectrum tiene la mayoria de metodos privados. Esto hace pensar en su utilidad.

Por que todo es privado?

Esto puede ser un code smell

Por favor leer: https://carlosschults.net/en/are-private-methods-a-code-smell/

Agregar documentacion a todos sus modulos, metodos y clases

Copiar de pyedra

[testenv:docstyle]
deps = pydocstyle
commands =
    pydocstyle nirdust.py --convention=numpy

For example:

noise of the spectra
noise_factor in line_spectrum method
difference in resolution between spectra
etc

Acordarse de agregarlo a travis

[testenv]
deps =
    ipdb
    pytest
commands =
    pytest test_nirdust.py {posargs}

Provide a way to allow the wavelength to be in different units, like micrometers, angstrom, nanometers, etc. Maybe automatically detect the units in "read_spectrum" according to the order of magnitude of the spectral axis.

compute_noise is doing something wrong. Either it is not computing the noise in the region asigned by the user or it is not resetting the "noise" parameter

• Primero definir el test
• Funcion de corte cut_edges

def test_cut_edges(NGC4945_continuum):
    spectrum = NGC4945_continuum
    reg = sp.SpectralRegion(min, max)
    expected = sp.manipulation.extract_region(spectrum.spec1d, region)
    result = spectrum.cut_edges(min, max)
    np.testing.assert_array_equal(result.spec1d.flux, expected.flux)

Remove the fits extension as argument. It should be detected automatically with other header cards.

Prototype

# ==============================================================================
# DOCS
# ==============================================================================

"""Nirdust test suite."""


# ==============================================================================
# IMPORTS
# ==============================================================================

import numpy as np
from astropy.io import fits
import matplotlib.pyplot as plt
import specutils as su
import os
import pathlib
from pathlib import Path
from astropy import units as u


# ==============================================================================
# CONSTANTS
# ==============================================================================

PATH = pathlib.Path(os.path.abspath(os.path.dirname(__file__)))


# ==============================================================================
# CLASSES
# ==============================================================================


class NuclearSpectrum(su.Spectrum1D):
    def __init__(
        self,
        
        flux,
        header, 
        spectrum_length,
        
        radius=None,
        spectral_axis=None,
        data=None,
        flux_unit=None,
        wav_unit=None,
        wcs=None,
        velocity_convention=None,
        rest_value=None,
        redshift=None,
        radial_velocity=None,
        bin_specification=None,
        uncertainty=None,
        mask=None,
        meta=None,
        dispersion_key="CRVAL1",
        wavelength="CD1_1",
    ):

        self.spectrum_length = spectrum_length
        self.radius = radius

        super().__init__(
            flux=self.spectrum[extension].data * u.adu,
            spectral_axis=(
                self.header[dispersion_key]
                + self.header[wavelength]
                * np.arange(0, spectrum_length)
            )
            * u.AA,
            wcs=wcs,
            velocity_convention=velocity_convention,
            rest_value=rest_value,
            redshift=redshift,
            radial_velocity=radial_velocity,
            bin_specification=bin_specification,
            uncertainty=uncertainty,
            mask=mask,
            meta=meta,
        )





# class OffNuclearSpectrum(NuclearSpectrum):
#     ...



# ==============================================================================
# LOAD SPECTRA
# ==============================================================================


def read_nuclear_spectrum(fname, extension, **kwargs):  # IO
    spectrums = []
    with fits.open(fname) spectrum:
        fluxx = pectrum[extension].data
        header = fits.getheader(spectrum)
        spectrum_length = len(spectrum[extension].data)

    single_spectrum = NuclearSpectrum(
        spectrum_length=spectrum_legth, 
        flux=flux, header=header, 
        *args, **kwargs)
        
    return single_spectrum


spectrums = []
for path in glob.glob("/spectrums*.fots"):
    spec = nd.read_nuclear_spectrum(path, 0)
    spectrums.append(spec)

delete "dispersion key", "first_wavelength", "dispersion_type"

spectrum_lenght must be updated after every operation

can other information be added?

Implement warning for unsuccesfull fit

• Configure the .travis.yml file
• Enable the continuous integration to the master branch
• Add the build badge to the README.md

Copy the .travis.yml file from grispy and modify it

El método tiene que ser un verbo normalize por ejemplo

Add two flake8 plugins in tox.ini for the style environment.

The plugins only need to be installed and then flake8 works.

The plugins are:

• flake8-import-order: Ensures this stantar https://www.python.org/dev/peps/pep-0008/#imports
• flake8-black: validating Python code style with the command line code formatting tool black

these two plugins will achieve a much cleaner code

The tox environment configuration must be:

[testenv:style]
skip_install = True
usedevelop = False
deps = 
    flake8
    flake8-import-order
    flake8-black
commands = flake8 setup.py nirdust.py test_nirdust.py {posargs}

Also to correctly configure the black tool you need to create the pyproject.toml file, with the content

[tool.black]
line-length = 79
target-version = ["py38"]

Implement it or remove it from the return statement until you do.

Las funciones no empiezan con mayuscula y son un verbo descriptivo.

LAs funciones tienen nombres tipo substract_spectrum_to_another_zaraza

La documentación no esta clara en su resumen principal

"Perform operations upon nuclear and external spactra." eso no dice nada

Modifie sp_correction to deal with different resolution/length spectra

Read the Docs is not building the documentation on every push. The problem is in the webpage configuration of read the docs.

Change custom_model to Fittable1DModel to allow parallelization.

i think this normalization could be more accurate. Check

• Pasar todos ipynb y el README a una carpeta draft
• nirdust.py <<< archivo con codigo relevante
• Agregar antes que nada el .gitignore
• test_nirdust.py
• Agregar LICENCE
• tox.ini py38, style, coverage (al 80%)
• Readme: Que operaciones vas a querer en nirdust,

class NirdustSpectrum(Spectrum1d):
    ...

spectre = NirdustSpectrum(t=...)

spectre.predict(...)

import constants outside function

read_table no corrige por redshift

en la funcion _filter_internal

La linea 73 reza if atribute.name.endswith("_"): no deberia ser:

if atribute.name.startswith("_") or atribute.name.endswith("_"): 
    ...

In this moment the folders where the spectra are stored and the name of the optional list of spectra are fixed. It would be good to allow the user to name the folders and the list as they wish.

Also, allow the user to input the unities

check why it is interpolating when cutting.

• Copy and adapt the doctstyle environment from grispy.
• Add the environment in the [tox] envlist.
• Add the new environment into travis.

It may be interesting to leave to the user the possibility of choosing the fitter.
See https://docs.astropy.org/en/stable/modeling/reference_api.html#module-astropy.modeling.fitting

Implement a class method for NirdustSpectrum that rebins or smooth spectra in order to increment the S/N. The imput could be the desired S/N (with some warning when the requeriment is too high) or the kernel to use in the smoothing.

In both read_table and read_fits z parameter should be None by default and a new parameter is needed to indicate if the spectrum must be redshift corrected

Inputs should accept astropy Quantity with units. The outputs should have units only if inputs have units.