Daycount fractions for minimal support on ratecurve

[disentchris]

Is your feature request related to a problem? Please describe. Part of roadmap Describe the solution you'd like In rate curve right now I overage Duration,yf and Duration.just_bds:

Duration.just_bds = lambda self, *args, **kwargs: self.just_days
Duration.yf = lambda self, *args, **kwargs: self.just_days / 365

Additional context Here's how I use them

(validated_date - validated_root).just_bds(cal=cal, dc=dc)
(validated_date2 - validated_date1).yf(cal, dc)

[amalizzio]

I will add the 3 simple ones:

ACT/365 (aka ACT/365F or ACT/365 FIXED where actuals / 365, not ACT/ACT ISDA where denominator counts days in/not in leap year) ACT/360 (actuals / 360) 30E/360 (min d1/d2 to 30, using the classic formula)

https://en.wikipedia.org/wiki/Day_count_convention

Future: should review Wikipedia against packages (like QL or Rateslib) and compare to source docs (eg ISDA spec...)

[amalizzio]

@Batu4 can you please pull my branch feat-daycounters

I have provided the code, i need you to validate the calculations

• understand the purpose of the 4 day counters in the new yfs.py (no more than 2hours google/gpt, there's a link in the funciton)
• review the calculations
• review the test cases in test_duration.py test_yfs
• install locally either quantlib-python or rateslib and perform those 4 examples, and put the results in expectdcf* variables as expected results as numbers
• spend 10-15 minutes thinking about adding other test cases should we be missing any edge cases on the code
• ensure coverage tests are 100%, ruff/black/isort, ...
• pull request my branch back to me, i was review again

[Batu4]

Started feat-daycounters-review:

• non-bd conventions - created a generator funcition, all passing
• BD/252 - issues, see attached spreadsheet validation (2 tabs NY and BR) - coming not ready yet

Todo

• for speed i commited quantlib and rateslib into the test_* files, but obviously they are not dependencies just testing, will remove them
  • book new issue for missing standard conventions like ACT/ACT ISDA

FYI - coverage/ruff/isort/black all 100% as expected

[amalizzio]

@Batu4

• Please do better formatting of issue comments here (read https://github.com/adam-p/markdown-here/wiki/Markdown-Cheatsheet), I have mentioned this several times before
• This topic is for day counting, not feat-durationstr_2nd, DO NOT MIX ISSUES
• A 3% tolerance on a $10MM swap is $300,000 (that equals jobs). No tolerance is acceptable for for daycounting. I expect you to understand that fully. Use quantlib as a REFERENCE only, get the exact answer. Then if they do not match I expect a spreadsheet attached with differences 100% attributed and highlighted for discussion.
• You said you would do more test cases, we discussed EOM and leap/non-leap -year scenarios scenarios, where are there?
• you need to check the BRL holidays in both and compare, and attach a spreadsheet of both to fine the exact differences
• I did a branch compare between feat-daycounters-review and feat-daycounters and there is no code change, what is the point of 2 branches (there is nothing to review) if you do the fix in branch where I did, we discussed yesterday, you spawn a new branch, then you do a pull request when complete, issues is not for your daily work review
• I compared feat-daycounters against dev and I can see your code changes (mixed withmine), you have add quantlib as a dependency to dateroll, we don't just add 3rd party dependencies to libraries without a face to face discussion
• Remove the 3rd party dependencies (rateslib and quantlib), they are for bench-marking, expected value should be an exact number, no imports are allowed in dateroll without a live discussion, I thought this is should be obvious as i said "put the results in expectdcf* variables as expected results as numbers" (emphasis on numbers)

[Batu4]

feat-daycounters-review

• removed 3rd party dependencies from tests and let the tests fully dependent on json files

[Batu4]

BD252 Issues: testcases report in the spreadsheet and I attached the spreadsheet here

Day_Count.xlsx

[Batu4]

Summary:

• In the testcases of spreadsheet, day conventions of act360,act365,30e360 of dateroll matches with the quantlib and excel calculation;however, bd252 with brazilian calendar does not match with quantlib because dateroll holidays did not include carnival,good friday and corpus christi

• Testcases include leap year, non leap year scenarios

• You are able to compare feat-daycounters-review and feat-daycounters branch now

• I would not be able to finish the excel calculation of bd252 in the spreadsheet

• Removed 3rd party dependencies from feat-daycounters-review branch

• Coverage 100%

Todo:

• Finish the bd252 excel calculation
• add carnival,good friday and corpus christi into the calendar of dateroll

[Batu4]

Summary:

• Last time, the dateroll brazilian holidays did not include carnival, good friday and corpus christi whereas quantlib included those holidays. I switched dateroll brazilian holidays to quantlib brazilian holidays. Now all day conventions of act360,act365,30e360 match with excel and quantlib, bd252 match with quantlib. (please see the attached excel file, sheet testcases)

• I removed the bd252 excel calculation in the sheet testcases because the holidays of the excel calculation will be coming from quantlib

• The branch name is feat-bd252-adjust and I did pull request to dev Day_Count_Complete.xlsx