Compare parfiles

[dlakaplan]

• Command-line utility to compare par files
• Use colors for output
• Bugfixes to reduce crashes

[dlakaplan]

Usage:

(pintdev) kaplan@plock[~/pythonpackages/PINT_dlakaplan] (comparepar) % compare_parfiles -h                         
usage: compare_parfiles [-h] [--dmx] [--no-dmx] [--sigma SIGMA] [--uncertainty_ratio UNCERTAINTY_RATIO] [--nocolor]
                        [--comparison {max,med,min,check}] [--log-level {TRACE,DEBUG,INFO,WARNING,ERROR}]
                        input1 input2

PINT tool to compare parfiles

positional arguments:
  input1                First input parfile
  input2                Second input parfile

optional arguments:
  -h, --help            show this help message and exit
  --dmx                 Print DMX parameters (default: False)
  --no-dmx              Do not print DMX parameters (default: False)
  --sigma SIGMA         Pulsar parameters for which diff_sigma > threshold will be printed with an exclamation point at the end of the line (default:
                        3)
  --uncertainty_ratio UNCERTAINTY_RATIO
                        Pulsar parameters for which the uncertainty has increased by a factor of unc_rat_threshold will be printed with an asterisk
                        at the end of the line (default: 1.05)
  --nocolor             Turn off colorized output (default: False)
  --comparison {max,med,min,check}
                        "max" - print all lines from both models whether they are fit or not (note that nodmx will override this); "med" - only print
                        lines for parameters that are fit; "min" - only print lines for fit parameters for which diff_sigma > threshold; "check" -
                        only print significant changes with logging.warning, not as string (note that all other modes will still print this)
                        (default: max)
  --log-level {TRACE,DEBUG,INFO,WARNING,ERROR}
                        Logging level (default: WARNING)

[dlakaplan]

@scottransom : I am trying to colorize the output. But I want to just adjust the background color while leaving the foreground static. is there a way to do that?

[scottransom]

Hmmm. Good question. I think to do that you would somehow have to query your terminal environment to know what the current foreground and background colors are. I certainly don't know of a way to do that.

[aarchiba]

Would it be possible to (optionally?) output a Markdown table, where markdown highlights (*...* or /.../ say) were used to indicate things of interest? A Markdown table is quite a clear ASCII representation but also allows a web browser to render a nicer version. In particular there are libraries to handle displaying Markdown nicely (with colours!) on a terminal: https://python.plainenglish.io/dump-the-plain-text-help-in-your-command-line-interfaces-today-3282225274dd

(ETA: And notebooks, of course, can natively display Markdown output with a very little persuasion.)

[dlakaplan]

I was part-way to turning it into more table-like output anyway, so if I finished that adding markdown would be trivial.

I did search on how to find the current foreground/background colors, but couldn't find a good answer. This is mostly relevant since e.g., I have white text on black as the default, so just fg=black isn't right.

[codecov]

Codecov Report

Merging #1340 (3039866) into master (24e144a) will decrease coverage by 0.06%.
The diff coverage is 59.82%.

@@            Coverage Diff             @@
##           master    #1340      +/-   ##
==========================================
- Coverage   61.31%   61.25%   -0.07%     
==========================================
  Files          89       90       +1     
  Lines       20218    20349     +131     
  Branches     3619     3646      +27     
==========================================
+ Hits        12397    12465      +68     
- Misses       7046     7097      +51     
- Partials      775      787      +12     

Impacted Files	Coverage Δ
src/pint/scripts/compare_parfiles.py	0.00% <0.00%> (ø)
src/pint/models/timing_model.py	81.63% <67.85%> (-1.77%)	⬇️
src/pint/utils.py	57.69% <100.00%> (+1.41%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 24e144a...3039866. Read the comment docs.

[dlakaplan]

Here is what the markdown output looks like:

[dlakaplan]

@almcewen0 : since you worked on this some, maybe you can take a look. Especially if you use this in notebooks see how the markdown output looks

[almcewen0]

i think it looks great on the command line! it doesn't run in a jupyter notebook, but that might be better suited for the model.compare() function.

[dlakaplan]

The command-line version wraps model.compare(). it's changes to that function which do markdown in a notebook

[dlakaplan]

@swiggumj : would this markdown output be using in timing analysis?

[swiggumj]

This looks great. Yeah, I think so!... I'd have to check, but I think it also might help clean up some of the code where things get compiled into per-pulsar summary files.

[dlakaplan]

Now included in https://nanograv-pint--1340.org.readthedocs.build/en/1340/examples/understanding_fitters.html#Comparing-models

[aarchiba]

This looks great. Yeah, I think so!... I'd have to check, but I think it also might help clean up some of the code where things get compiled into per-pulsar summary files.

Indeed. This could (should!) be used directly in the timing_analysis notebooks so you can show the comparison. For the summaries, right now the table is shown as preformatted, but this would allow us to produce much nicer-looking tables as a drop-in replacement.

[dlakaplan]

Were you thinking that the fit.get_summary() method should also return markdown? Possible - I just didn't know if it would be useful. But I can add it.

[scottransom]

That looks sweet. But looking at the readthedocs example, what is up with F1?!

[dlakaplan]

No idea - it's also in the cells above when the Powell fit results are shown. And I think that has been around for a while:
https://nanograv-pint.readthedocs.io/en/latest/examples/understanding_fitters.html#Powell-fitter
So I can investigate but probably a separate issue.

[dlakaplan]

@scottransom : actually wait a second. I want to update the docs slightly.

[scottransom]

Oops! I saw that too late!

[dlakaplan]

OK, I guess I was too late. I will just edit that directly then (just to note that without raw=True the markdown in notebooks will vanish)

[scottransom]

I think you can just merge the doc changes directly in since they don't affect our CI tests