brandon-rhodes
commented
6 months ago I would guess that for P type orbit the packed designation should be nan but in this case it reads the fragment part.
Let's read down through the whole file to see how things look. This little command line:
cat CometEls.txt | awk '{print substr($0,1,24)}'—shows that the file is cleanly split between lines with designations in one format, and a second section that uses another. Omitting long stretches of lines with the same format, here's what we see:
CJ95O010 1997 03 29
PJ96R020 2019 02 12
PJ98V24S 2018 01 19
...
CK22D020 2022 03 27
PK22E010 2022 10 10
CK22E020 2024 09 13
CK22E030 2023 01 12
CK22F010 2022 07 28
CK22F020 2022 03 23
CK22H010 2024 01 18
CK22J010 2022 02 16
0001P 1986 02 17
0002P 2023 10 20
0004P 2021 09 8
...
0050P 2024 05 12
0051P a 2022 10 3
0051P d 2022 10 1
0052P 2021 10 5
...
0441P 2025 09 7
0001I 2017 09 9
0002I 2019 12 8The first thing we can note is that we can't, alas, use your suggestion of using the orbit type P to decide between the two formats, because there are P-type comets in both sections.
The second thing we note is that, as you correctly note, the same field is used sometimes for the (mostly blank) fragment number (I included comet 0051P specifically so we'd get to see an example of a case where the field wasn't blank), and sometimes it's used for a modern packed designation like the ones used for minor planets.
For reference, here's how the format is described on their website:
https://www.minorplanetcenter.net/iau/info/CometOrbitFormat.html
Columns F77 Use
1 - 4 i4 Periodic comet number
5 a1 Orbit type (generally `C', `P' or `D')
6 - 12 a7 Provisional designation (in [packed form](https://www.minorplanetcenter.net/iau/info/PackedDes.html))Which makes it sound so simple! But over in their text description of the format, they describe two different ways the fields can be used, with the "packed form" designation instead being a fragment number:
https://www.minorplanetcenter.net/iau/info/PackedDes.html
So the problem here is that Pandas doesn't allow columns to have different names depending on their content. There's no way (at least that I know of?) to tell Pandas, "the third column of this file should be called 'fragment number' when there's a non-empty comet number in columns 1-4, but should be called 'packed designation' when columns 1-4 are blank."
So in my design for the column numbers, I tried to just stick with the "official" names of the columns given in the first link above.
Before wading into further details, let's step back a moment. When I wrote up the code to handle comments, I didn't pay attention to those first columns, because the lines of the file are now much longer than in the old days, and include the full text of the comet name way over to the right. In your example, we can just read the full name 323P-B/SOHO from the big field way over to the right, without needing to wade into the fields on the left. I guess the issue you've encountered is that the full name does not include the fragment number, the way the MPC have designed that new full-name field?
So what you want is a way to find the fragment number. Hmm.
Maybe the Skyfield documentation on comets could include a little snippet showing how to get the fragment number? It might work something like this:
import io
from skyfield.data import mpc
for comet_orbit in [
# Three example lines from CometEls.txt of May 10 2022:
' CJ95O010 1997 03 29.6937 0.892238 0.994969 130.4267 282.8976 89.2092 20220510 -2.0 4.0 C/1995 O1 (Hale-Bopp) MPC106342',
' PJ98V24S 2018 01 19.1688 3.427137 0.242855 244.6040 159.0401 5.0266 20220510 13.0 2.0 P/1998 VS24 (LINEAR) MPC 75703',
'0343P 2017 01 29.8032 2.266196 0.584845 137.4257 257.1514 5.5878 20220510 14.0 4.0 343P/NEAT-LONEOS MPC109148',
# Your example line:
'0323P b 2025 12 16.3240 0.040025 0.986146 353.9756 323.4582 5.4670 20240331 26.0 4.0 323P-B/SOHO MPEC 2024-F21',
]:
orbit = io.BytesIO(bytes(comet_orbit, 'utf-8'))
c = mpc.load_comets_dataframe_slow(io.BufferedReader(orbit))
c = (c.groupby('designation', as_index=False).last()
.set_index('designation', drop=False))
row = c.iloc[0]
number = row.number
if number > 0:
fragment_number = row.designation_packed
else:
fragment_number = row.designation_packed[6].rstrip('0')
print('Orbit type {!r} Designation {!r} Fragment number {!r}'
.format(row.orbit_type, row.designation, fragment_number))Hmm. There's a complication here. Because we're only giving one line at a time, Pandas doesn't have the full CometEls.txt to examine, and has to guess about column data types based on only one example line. Thus the output of my little example:
Orbit type 'C' Designation 'C/1995 O1 (Hale-Bopp)' Fragment number ''
Orbit type 'P' Designation 'P/1998 VS24 (LINEAR)' Fragment number 'S'
Orbit type 'P' Designation '343P/NEAT-LONEOS' Fragment number nan
Orbit type 'P' Designation '323P-B/SOHO' Fragment number 'b'— gets a fragment number of nan, a floating point value, where we really want '' the empty string, because Pandas is guessing that the blank column is a float rather than a string. I guess we could spruce up the import a bit to know more about types.
But, first, I'll stop here to hear your comments, to learn whether getting the fragment value like this would answer your purpose — or whether there's some other underlying goal?
jurezakrajsek
I noticed that for some comets the packed designation is parsed wrong This is the MPC 1-line orbital elements beeing parsed
0323P b 2025 12 16.3240 0.040025 0.986146 353.9756 323.4582 5.4670 20240331 26.0 4.0 323P-B/SOHO MPEC 2024-F21And the output of the mpc.load_comets_dataframe_slow(io.BufferedReader(orbit))
I would guess that for P type orbit the packed designation should be nan but in this case it reads the fragment part.
Here is a short script to reproduce: