empty table cells in multi-period balance reports [since 1.20]

$ hledger bal -f tmp.journal --yearly Income:Gains
Balance changes in 2018-01-01..2021-12-31:

              || 2018  2019  2020  2021
==============++========================
 Income:Gains ||    0                 0
--------------++------------------------
              ||    0                 0
$ hledger bal -f tmp.journal -p 2019 Income:Gains
-93.56707511000000000000000000 EUR  Income:Gains
--------------------
-93.56707511000000000000000000 EUR
$ hledger bal -f tmp.journal -p 2020 Income:Gains
-1094.29560548000000000000000000 EUR  Income:Gains
--------------------
-1094.29560548000000000000000000 EUR

I don't understand what is going on here. The --yearly report has always worked for me, and that stopped recently.
Now I cannot decide whether my understanding is wrong, or hledger has regressed. In any case, it is surprising.

Forgot to mention that it's on 1.21.

Thanks for the report. That's unfortunate! Would you be able to boil it down to an example journal ?

Thanks for the report. That's unfortunate! Would you be able to boil it down to an example journal ?

I'll try, but In the event that I'm unsuccessful, will you have a way to receive the tmp.journal securely?
gpg, onionshare, keybase, whatever...

I use a lot of piped hledger print -f - commands, and commodity directives don't seem to be preserved in the output of those print commands. So tmp.journal does not have a commodity directive. Adding it back makes everything magically work.

(echo "commodity 1000.00 EUR"; cat tmp.journal) | hledger bal -f - --yearly Income:Gains
Balance changes in 2018-01-01..2021-12-31:

              || 2018        2019          2020  2021
==============++======================================
 Income:Gains ||    0  -93.57 EUR  -1094.30 EUR     0
--------------++--------------------------------------
              ||    0  -93.57 EUR  -1094.30 EUR     0

So the failure mode seems formatting related.

There's a rather large count of digits after the decimal point for most amounts in the tmp.journal, since it came from unrealized gains calculations computed in beancount.

$ grep -oE "[0-9]+\.[0-9]+ EUR" tmp.journal | cut -d ' ' -f 1 | cut -d '.' -f 2 | awk '{print(length)}' | sort -rn  | uniq -c
      2 29
      5 28
     46 27
      4 26
      2 25
    273 24
      6 23
    547 10
   2789 8
    194 4

Sure, if you have a matrix client set up, encrypted direct chat to me (simonmic:matrix.org) could be an easy way.

I'm curious: do you still get this issue in master, or with the branch in #1509? I've put some effort into refactoring the code so this sort of bug (easy to accidentally do in the past) is less likely/impossible to do now.

@Xitian9

I'm curious: do you still get this issue in master, or with the branch in #1509? I've put some effort into refactoring the code so this sort of bug (easy to accidentally do in the past) is less likely/impossible to do now.

@Xitian9 I will check today, and let you know

@Xitian9

@Xitian9 tried your mixedamt branch. Same problem. I'm going to try to shrink the journal, and if not, i'll need to setup a matrix client to send @simonmichael the journal.

Here is the smallest journal that exhibits the problem on my end:

cat <<EOF | ~/.local/bin/hledger-1.21 bal -f - --yearly Income:Gains

2020-12-30 * Sold BTC to EUR
  Income:Gains                                                      -0.08724382 EUR
  Assets:XBT                                                       -0.00000437 XBT @ 2919.034429281334766023231553 EUR
  Assets:EUR                                                         0.10000000 EUR
  Equity:Rounding  ; rounding from beancount: 0.0000000004559594329275215219 EUR

2021-04-11 U Unrealized loss for 14.23000000 units of GBP (price: 1.1526 EUR as of 2021-04-09, average cost: 1.1680 EUR)
  Assets:GBP:Unrealized                                            -0.21871744220000000000000000 EUR
  Income:GBP:Unrealized                                            0.21871744220000000000000000 EUR
  Equity:Rounding  ; rounding for hledger

EOF

Results in

Balance changes in 2020-01-01..2021-12-31:

              || 2020  2021
==============++============
 Income:Gains ||          0
--------------++------------
              ||          0

@Xitian9

Thanks!

@Xitian9, I think showAmountsOneLineB shows nothing if the rendered amount would exceed the hard-coded max width of 32 in Hledger.Cli.Commands.Balance.balanceOpts. (In this example, writing 25 instead of 26 decimal places makes it work.) It should show an elided amount instead, I guess.

Or, allow cells to grow as wide as needed ?

[Start of a test:]

<
2021-01-01
    (a)      -0.12345678901234567890123456 EUR

$ hledger -f- bal -YN
Balance changes in 2021:

   ||                              2021 
===++===================================
 a || -0.12345678901234567890123456 EUR

It's not clear how to elide a single amount. Here are some options:

Elide from the left
- Simple
- May elide commodity symbol
- Elides most significant digits
Elide from the right
- Simple
- May elide commodity symbol
- If there is no visible decimal place, cannot determine the order of magnitude of the number
Elide from the left, but don't elide commodity symbol
- Similar to (1), but doesn't elide commodity symbol.
- Still runs into problems with long commodity symbols.
Elide from the right, but don't elide commodity symbol
- Similar to (2), but doesn't elide commodity symbol.
- Still runs into problems with long commodity symbols.
Use scientific notation for elided numbers
- Number is clear
- Still runs into problems with long commodity symbols

Or, as you say, we can just ignore the maximum width if it would cause nothing to display. This might mess up alignment unless we put some extra effort into that.

I was imagining just whatever kind of eliding we already do with too-wide amounts, even it it's not yet ideal it would be consistent. But what did 1.19 do - expand the column without limit to fit the widest amount ? That's not always super usable either but for now it's probably the most useful. Eliding can be bothersome.

We currently do eliding by displaying only some of the amounts in a MixedAmount; we don't ever elide the string representing a single amount.

In 1.19, exactly two quantities would be displayed, with no regard for their maximum width.

So perhaps the best option is to elide according to the previous rules, but make sure we always display at least one commodity along with an elision string, even if that would put us over the hard-coded width. That width is pretty arbitrary in any case.

lib: When displaying amounts on a single line, always display at leas…

Sounds pretty good..

lib: When displaying amounts on a single line, always display at leas…

Fixed in master by #1530.

I can confirm that it is fixed in master 0f4e462. Thanks!

Navigation Menu

Repository navigation

empty table cells in multi-period balance reports [since 1.20] #1526

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

empty table cells in multi-period balance reports [since 1.20] #1526

Description

Activity

lestephane commented on Apr 11, 2021

simonmichael commented on Apr 11, 2021

lestephane commented on Apr 11, 2021

lestephane commented on Apr 11, 2021

lestephane commented on Apr 11, 2021

simonmichael commented on Apr 11, 2021

Xitian9 commented on Apr 12, 2021

lestephane commented on Apr 13, 2021

lestephane commented on Apr 13, 2021

lestephane commented on Apr 13, 2021

simonmichael commented on Apr 13, 2021

simonmichael commented on Apr 13, 2021

Xitian9 commented on Apr 14, 2021

simonmichael commented on Apr 14, 2021

Xitian9 commented on Apr 14, 2021

Xitian9 commented on Apr 14, 2021

simonmichael commented on Apr 14, 2021

simonmichael commented on Apr 14, 2021

lestephane commented on Apr 14, 2021

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions