Go port of ledger-cli WAS Re: Abridged summary of ledger-cli@googlegroups.com - 3 updates in 2 topics

[Daryl Manning]

On Sat, Nov 1, 2025 at 7:19 AM <ledge...@googlegroups.com> wrote:

ledge...@googlegroups.com

Oh, this is super helpful.Thanks!

Yeah, I was trying to understand how I got the 20k lines of C++ code
down to about 1500 lines of Go for the core, but imagine I will constrain the format much much more tightly
and avoid some of the more "turing complete" options available. It is impressive what's been done, but
think I am focusing on a data format more than a language itself.

I've got an initial, naive implementation done which parses and gives balances register etc etc for my personal
accounts, but noticed a few issues and need to throw it against some more complex data for things like
my stock portfolio and business accounts.

I'll post up the Go implementation shortly. I used regexs ofr prasing (which I know are more fragile) but also
think I can move to some sort of AST later as it develops. Was kinda fun though. I am still trying to grok a
lot of the niche dege cases in the code base and options which (tbh) I think I'll never use.

Anyhow, thanks for the reply!

 

Simon Michael wrote

On 2025-10-29 17:27, Daryl Manning wrote:
> I'd be interested in hearing about the variability you think there is in
> the file format (I use a stricter format for *myself*)
> like, for example, not allowing prefixed currencies and such and
> demanding ISO 3 letter curency codes etc, but would be interested
> in hearing the types of things that we all feel _should_ be removed.

It's very complex. A few things that come to mind (still true as far as
I know):

- Multiple formats can be mixed in the journal file. Journal entries and
directives, but also timeclock entries, python code and value
expressions (as part of journal entries and directives I think) which
are a turing complete programming language themselves.

- There's great flexibility in the syntax, which is a double edged
sword. It feels pleasant for users in the beginning but complicates life
ever after for implementors (and in the end, for users as well).
Beancount's syntax is too restrictive for comfort in some ways, but eg
requiring strict ISO dates, and ISO currency codes on the right hand
side, are two small things that simplify a lot.

- The syntax for everything right of the account name in postings stands
out as particularly complex - amounts, costs, lot prices, lot dates, lot
labels, balance assertions, balance assignments, and comments can all be
combined there, in many different ways.

- Allowing spaces in account names brings in the two-space delimiter
requirement, which beginners always trip up on. Also it makes scripting
harder, later. Again it's a double-edged sword, costly though also
sometimes pleasant.

You might be interested to review hledger's journal format[1]. It copies
most though not all of Ledger's (not value expressions, eg). But you can
see I've documented some syntax as first class and pushed some other
parts into a legacy section[2]. Also I discuss syntax differences at [3].

[1] https://hledger.org/1.50/hledger.html#journal
[2] https://hledger.org/1.50/hledger.html#other-syntax
[3] https://hledger.org/ledger.html#journal-format-differences

You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page. To unsubscribe from this group and stop receiving emails from it send an email to ledger-cli+...@googlegroups.com.