csv import: Fails to import "Long_Month unpadded_date, Year" into txn date

[sten0]

hledger 1.52.1-g3834a163b-20260428, linux-x86_64

 "June 30, 2026" as a date using date format "\"%m %d, %Y\""
record: June 30, 2026,Action,Symbol,Long Name,,0.05,June 30, 2026,0000001,1.00,COIN,Long Description for Note Field
the date rule is:   %1
the date-format is: "%m %d, %Y"
you may need to change your date rule, change your date-format rule, or change your skip rule
for m/d/y or d/m/y dates, use date-format %-m/%-d/%Y or date-format %-d/%-m/%Y

The actual record is:

"June 30, 2026",Action,Symbol,Long Name,,0.05,"June 30, 2026",0000001,1.00,COIN,Long Description for Note Field

I've tried every combination I can think of for the date-format rule. This bug overlaps with the internationalisation issue we're also discussing.

I expected heldger to import this RFC 4180-compliant record while preserving the quoted string, pass it through iso8601-time or similar to do the equivalent of echo "June 30, 2026" | date -I. It wasn't clear to me in the documentation whether or not hledger can do this yet.

[sten0]

Clearly I'm half-blind and didn't see a blue link on a blue background (https://hackage-content.haskell.org/package/time-1.16.0.1/docs/Data-Time-Format.html#v:formatTime), but I later found it while rereading the man page. Currently we have

              # MM/DD/YY
              date-format %m/%d/%y

              # D/M/YYYY
              # The - makes leading zeros optional.
              date-format %-d/%-m/%Y

              # YYYY-Mmm-DD
              date-format %Y-%h-%d

and I'd like to make a PR that makes the docs easier to understand, but only adds a few words (maybe 8 words more, tops?). I think it's also important to specify that %b and %B are localised month names. What is the rationale behind choosing %h instead of %b, by the way? %h is irregular when %Y (long year), %y (short year), %B (long month), %b (short month). %h also seems obscure when date uses

       %b     locale's abbreviated month name (e.g., Jan)

       %B     locale's full month name (e.g., January)

Yes, I'm mostly running into documentation issues caused by "paradox of knowledge", and the reason I'm taking the time to report (and soon to file a PR for) these issues is to make onboarding easier and smoother.

[simonmichael]

to make onboarding easier and smoother.

That's excellent, thank you!

I hope to catch up on this soon.

[simonmichael]

Clearly I'm half-blind and didn't see a blue link on a blue background (https://hackage-content.haskell.org/package/time-1.16.0.1/docs/Data-Time-Format.html#v:formatTime), but I later found it while rereading the man page.

That's interesting. I guess you were on the website, but using one of the dark themes ? I usually see it like this:

[simonmichael]

Whereas Navy theme looks like this (for me):

[simonmichael]

I think it's also important to specify that %b and %B are localised month names. What is the rationale behind choosing %h instead of %b, by the way?

That does sound important! I wasn't aware they were different, the haddock docs are quite cryptic:

%b, %h
month name, short form (snd from months locale), Jan - Dec

[simonmichael]

What action(s) are needed here ?

[sten0]

Hi Simon, Simon Michael ***@***.***> writes:

simonmichael left a comment (simonmichael/hledger#2668) What action(s) are needed here ?

There are two issues here. I'm currently in the process of learning about upstream localised month names. Thus far it looks like Haskell is "multilingual and localised" for a set of languages that equals en_US, and I'm still in the process of determining if that's currently a bounded set. The veracity of my documentation PR will depend on upstream month (and possibly named day of week) locale support. Here is the second issue, and I'd appreciate if you would work on this since it doesn't appear to depend on upstream localised time/date support: When I edit my bank-provided csv to delete the comma in the quoted date field, the rule date-format %B %-d %Y functions correctly; however, when I do not remove the comma, the rule date-format %B %-d, %Y fails with the error I originally reported. The minimal reproducer is an RFC 4180 CSV record like this: "Date",Action,Symbol,Description,shares,price "July 20, 2026",Sell,AAA,Long Description,10,50 I'm now using an external tool to validate CSV input to be 100% certain that hledger (or haskell upstream!) is at fault ;) Best, Nicholas

[simonmichael]

Hi Nicholas, I can't reproduce your second issue. This works for me as expected:

~/src/hledger$ cat a.csv
"Date",Action,Symbol,Description,shares,price
"July 20, 2026",Sell,AAA,Long Description,10,50
~/src/hledger$ cat a.csv.rules
skip 1
fields date, action, symbol, description, shares, price
date-format %B %-d, %Y
~/src/hledger$ hledger -f a.csv print
2026-07-20 Long Description

and with the comma removed from date-format, it fails as expected:

hledger: Error: could not parse "July 20, 2026" as a date using date format "%B %-d %Y"