Re: Debugging assertion failures / printing the value of a value expression

[John Lee]

Thanks Simon

I was referring to the combination of requirements below, which seems to
not quite fit ledger.  I don't mind so much how I achieve these with
particular features:

* Keep track of account balances (specifically bank balance)
intermittently without recording all transactions (either with balance
assignments or ordinary adjustment transactions that just add or
subtract an amount to reconcile with what the bank says is the current
balance)
* Periodically assert total account balance
* Track money set aside money from that account as e.g. savings

On Sat, 21 Oct 2017, at 18:05, Simon Michael wrote:
> On 10/20/17 1:46 PM, John Lee wrote:
> > I guess I should either dust off my very dusty C++ and try to fix it, or
> > try one of the ledger spinoffs like beancount... maybe there are
> > beancount / hledger people here who can comment re whether those systems
> > have ways to tackle the problem from the thread linked above?
> 
> Hi John,
> 
> I think you're referring to the fact that Ledger's balance assertions 
> check the exclusive balance only ? (excluding any subaccount balances).
> 
> hledger's balance assertions do the same, for compatibility. Sometimes 
> you want that (though I forget why ?). Sometimes it is a pain. We have 
> talked about adding new balance assertion syntax to check inclusive 
> balances. https://github.com/simonmichael/hledger/issues/290
> 
> There's also the issue of whether to check just the commodity of the 
> asserted amount, or the whole multicommodity balance of the asserted 
> account. Ledger and hledger currently do the former. 
> https://github.com/simonmichael/hledger/issues/195
> 
> 
> 
> -- 
> 
> --- 
> You received this message because you are subscribed to the Google Groups
> "Ledger" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to ledger-cli+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"Ledger" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to ledger-cli+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Debugging assertion failures / printing the value of a value expression

[Simon Michael]

On 10/20/17 1:46 PM, John Lee wrote:

I guess I should either dust off my very dusty C++ and try to fix it, or
try one of the ledger spinoffs like beancount... maybe there are
beancount / hledger people here who can comment re whether those systems
have ways to tackle the problem from the thread linked above?


Hi John,

I think you're referring to the fact that Ledger's balance assertions 
check the exclusive balance only ? (excluding any subaccount balances).


hledger's balance assertions do the same, for compatibility. Sometimes 
you want that (though I forget why ?). Sometimes it is a pain. We have 
talked about adding new balance assertion syntax to check inclusive 
balances. https://github.com/simonmichael/hledger/issues/290


There's also the issue of whether to check just the commodity of the 
asserted amount, or the whole multicommodity balance of the asserted 
account. Ledger and hledger currently do the former. 
https://github.com/simonmichael/hledger/issues/195




--

--- 
You received this message because you are subscribed to the Google Groups "Ledger" group.

To unsubscribe from this group and stop receiving emails from it, send an email 
to ledger-cli+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Debugging assertion failures / printing the value of a value expression

[John Lee]

Thanks for this, but I wonder if you have read the previous thread I
linked to fully?  It discusses what I believe (from admittedly only a
quick scan) are the same techniques you describe here, and the problems
I ran into with those.

Or did I miss something?

Thanks again, I do appreciate any alternative approaches, it seems quite
likely I've missed a better way.

On Fri, 20 Oct 2017, at 23:58, Michael Cooper wrote:
> Ah, I misunderstood your original problem, apologies.
> 
> Dodging your question again (since I really don't know the answer), I've
> also found that storing values in non-leaf accounts gets me into all
> sorts
> of trouble. It generally is confusing and my automation doesn't
> understand
> it. If this example were in accounts, I would do this:
> 
> 2017-10-01 * Opening Balance
> Assets:CurrentAccount:Real  $1000
> Equity:OpeningBalance
> 
> 2017-10-02 * Savings
> Assets:CurrentAccounts:Savings  $100
> Assets:CurrentAccounts:Real
> 
> This works with the "assert account(...).total" style of assertion, but
> totally fails with the Reconcile method I posted (it thinks that there is
> $0 in Assets:CurrentAccounts). The assert style still doesn't show you in
> what direction it is wrong, unfortunately.
> 
> Alternatively, this sounds sort of like the envelope-budgeting system I
> use. Would something like the below be interesting for you?
> 
> 2017-10-01 * Opening Balance
> Assets:CurrentAccount$1000
> (Budgets:SpendingMoney)  $1000
> Equity:OpeningBalance
> 
> 2017-10-02 * Savings
> [Budgets:Saved]  $100
> [Budgets:SpendingMoney]
> 
> 2017-10-03 * Reconcile
> [Assets:CurrentAccount]  $0 = $1000
> 
> The "Budgets" subtree are accounts made entirely of virtual transactions
> to
> track how money has been allocated, whereas the Assets subtree is about
> where money really is. I have scripts that help make sure that in
> transactions like 2017-10-01 the amount that the asset increases by
> balances with the amount the budgets increase by. I then have rules to
> automatically pair up Expense and Budget transactions.
> 
> On Fri, Oct 20, 2017 at 1:46 PM John Lee  wrote:
> 
> > Hi Michael
> >
> > If you take a look at my previous thread here from a week or so back
> > you'll see the problem you run into when trying to apply that to the
> > transactions I posted in this thread.
> >
> > https://groups.google.com/forum/#!topic/ledger-cli/bqu-qNr6cjM
> >
> > Consider:
> >
> > 2017-10-01 * Opening Balance
> > Assets:CurrentAccount   $1000
> > Equity:OpeningBalance
> >
> > 2017-10-02 * Savings
> > Assets:CurrentAccount:Savings   $100
> > Assets:CurrentAccount
> >
> > 2017-10-03 * Assertion
> > [Assets:CurrentAccount]   $0 = $1000
> >
> >
> > That final balance assertion fails for the same reason I mentioned in my
> > second post in the earlier thread:  that assertion refers not to the
> > summed-up total of everything in *or under* Assets:CurrentAccount (as
> > you'd see in the output of ledger b Assets:CurrentAccount), but the
> > amount that lives in that *exact* location (which you might write in a
> > balance command as
> > '^Assets:CurrentAccount$').  So $1000 errors, and $900 does not.  That
> > means I can't record the total balance of that real account, which is
> > what I'm after.  That's what motivated my use of assertion value
> > expressions, with which I guess I've either run into a bug, or I don't
> > understand something (see my previous post in this thread).
> >
> > I get the impression that transferring between subaccounts in this way
> > suggested here earlier, is a bit of a dark corner of ledger that's not
> > terribly well supported?
> >
> > I'm keen to hear about other approaches, but so far I'm running into
> > roadblocks whichever way I try to tackle using ledger the problem I set
> > out in that first thread (which is the same basic problem I'm trying to
> > solve here in this thread, though my question here was about a detail).
> >
> > I guess I should either dust off my very dusty C++ and try to fix it, or
> > try one of the ledger spinoffs like beancount... maybe there are
> > beancount / hledger people here who can comment re whether those systems
> > have ways to tackle the problem from the thread linked above?
> >
> > On Thu, 19 Oct 2017, at 23:54, Michael Cooper wrote:
> > > This isn't a direct answer to your question, but this is how I do these
> > > kind of balance checks:
> > >
> > > 2017/01/01 * Opening Balance
> > > Assets:Checking$1000
> > > Equity:Opening
> > >
> > > 2017/01/02 * Savings
> > > Assets:Savings  $100
> > > Assets:Checking
> > >
> > > 2017/01/03 * Reconcile
> > > [Assets:Checking]  $0 = $900
> > > [Assets:Savings]   $0 = $100
> > >
> > > 2017/01/04 * Savings
> > > Assets:Savings  $100
> > > Assets:Checking
> > >
> > > 2017/01/05 * Reconcile
> > > [Assets:Checking]  $0 = $800
> > > [Assets:Savings]   $0 = $200
> > >
> > > This 

Re: Debugging assertion failures / printing the value of a value expression

[Michael Cooper]

Ah, I misunderstood your original problem, apologies.

Dodging your question again (since I really don't know the answer), I've
also found that storing values in non-leaf accounts gets me into all sorts
of trouble. It generally is confusing and my automation doesn't understand
it. If this example were in accounts, I would do this:

2017-10-01 * Opening Balance
Assets:CurrentAccount:Real  $1000
Equity:OpeningBalance

2017-10-02 * Savings
Assets:CurrentAccounts:Savings  $100
Assets:CurrentAccounts:Real

This works with the "assert account(...).total" style of assertion, but
totally fails with the Reconcile method I posted (it thinks that there is
$0 in Assets:CurrentAccounts). The assert style still doesn't show you in
what direction it is wrong, unfortunately.

Alternatively, this sounds sort of like the envelope-budgeting system I
use. Would something like the below be interesting for you?

2017-10-01 * Opening Balance
Assets:CurrentAccount$1000
(Budgets:SpendingMoney)  $1000
Equity:OpeningBalance

2017-10-02 * Savings
[Budgets:Saved]  $100
[Budgets:SpendingMoney]

2017-10-03 * Reconcile
[Assets:CurrentAccount]  $0 = $1000

The "Budgets" subtree are accounts made entirely of virtual transactions to
track how money has been allocated, whereas the Assets subtree is about
where money really is. I have scripts that help make sure that in
transactions like 2017-10-01 the amount that the asset increases by
balances with the amount the budgets increase by. I then have rules to
automatically pair up Expense and Budget transactions.

On Fri, Oct 20, 2017 at 1:46 PM John Lee  wrote:

> Hi Michael
>
> If you take a look at my previous thread here from a week or so back
> you'll see the problem you run into when trying to apply that to the
> transactions I posted in this thread.
>
> https://groups.google.com/forum/#!topic/ledger-cli/bqu-qNr6cjM
>
> Consider:
>
> 2017-10-01 * Opening Balance
> Assets:CurrentAccount   $1000
> Equity:OpeningBalance
>
> 2017-10-02 * Savings
> Assets:CurrentAccount:Savings   $100
> Assets:CurrentAccount
>
> 2017-10-03 * Assertion
> [Assets:CurrentAccount]   $0 = $1000
>
>
> That final balance assertion fails for the same reason I mentioned in my
> second post in the earlier thread:  that assertion refers not to the
> summed-up total of everything in *or under* Assets:CurrentAccount (as
> you'd see in the output of ledger b Assets:CurrentAccount), but the
> amount that lives in that *exact* location (which you might write in a
> balance command as
> '^Assets:CurrentAccount$').  So $1000 errors, and $900 does not.  That
> means I can't record the total balance of that real account, which is
> what I'm after.  That's what motivated my use of assertion value
> expressions, with which I guess I've either run into a bug, or I don't
> understand something (see my previous post in this thread).
>
> I get the impression that transferring between subaccounts in this way
> suggested here earlier, is a bit of a dark corner of ledger that's not
> terribly well supported?
>
> I'm keen to hear about other approaches, but so far I'm running into
> roadblocks whichever way I try to tackle using ledger the problem I set
> out in that first thread (which is the same basic problem I'm trying to
> solve here in this thread, though my question here was about a detail).
>
> I guess I should either dust off my very dusty C++ and try to fix it, or
> try one of the ledger spinoffs like beancount... maybe there are
> beancount / hledger people here who can comment re whether those systems
> have ways to tackle the problem from the thread linked above?
>
> On Thu, 19 Oct 2017, at 23:54, Michael Cooper wrote:
> > This isn't a direct answer to your question, but this is how I do these
> > kind of balance checks:
> >
> > 2017/01/01 * Opening Balance
> > Assets:Checking$1000
> > Equity:Opening
> >
> > 2017/01/02 * Savings
> > Assets:Savings  $100
> > Assets:Checking
> >
> > 2017/01/03 * Reconcile
> > [Assets:Checking]  $0 = $900
> > [Assets:Savings]   $0 = $100
> >
> > 2017/01/04 * Savings
> > Assets:Savings  $100
> > Assets:Checking
> >
> > 2017/01/05 * Reconcile
> > [Assets:Checking]  $0 = $800
> > [Assets:Savings]   $0 = $200
> >
> > This works well. I read these reconcile transactions as "If $0 is added
> > to
> > the account, the balance should be $800".
> >
> > If I instead change the last reconcile to "[Assets:Checking]  $0 = $700"
> > (that is, make it $100 off) I get this error:
> >
> > While parsing file "/home/mythmon/tmp/foo.ledger", line 18:
> > While parsing posting:
> >   [Assets:Checking]  $0 = $700
> >   
> > Error: Balance assertion off by $-100 (expected to see $800)
> >
> > This shows me both the value I said it should be ($700), the value Ledger
> > says it should be ($800), and the difference ($-100). All of these end up
> > being very 

Re: Debugging assertion failures / printing the value of a value expression

[John Lee]

Hi Michael

If you take a look at my previous thread here from a week or so back
you'll see the problem you run into when trying to apply that to the
transactions I posted in this thread.

https://groups.google.com/forum/#!topic/ledger-cli/bqu-qNr6cjM

Consider:

2017-10-01 * Opening Balance
Assets:CurrentAccount   $1000
Equity:OpeningBalance

2017-10-02 * Savings
Assets:CurrentAccount:Savings   $100
Assets:CurrentAccount

2017-10-03 * Assertion
[Assets:CurrentAccount]   $0 = $1000


That final balance assertion fails for the same reason I mentioned in my
second post in the earlier thread:  that assertion refers not to the
summed-up total of everything in *or under* Assets:CurrentAccount (as
you'd see in the output of ledger b Assets:CurrentAccount), but the
amount that lives in that *exact* location (which you might write in a
balance command as
'^Assets:CurrentAccount$').  So $1000 errors, and $900 does not.  That
means I can't record the total balance of that real account, which is
what I'm after.  That's what motivated my use of assertion value
expressions, with which I guess I've either run into a bug, or I don't
understand something (see my previous post in this thread).

I get the impression that transferring between subaccounts in this way
suggested here earlier, is a bit of a dark corner of ledger that's not
terribly well supported?

I'm keen to hear about other approaches, but so far I'm running into
roadblocks whichever way I try to tackle using ledger the problem I set
out in that first thread (which is the same basic problem I'm trying to
solve here in this thread, though my question here was about a detail).

I guess I should either dust off my very dusty C++ and try to fix it, or
try one of the ledger spinoffs like beancount... maybe there are
beancount / hledger people here who can comment re whether those systems
have ways to tackle the problem from the thread linked above?

On Thu, 19 Oct 2017, at 23:54, Michael Cooper wrote:
> This isn't a direct answer to your question, but this is how I do these
> kind of balance checks:
> 
> 2017/01/01 * Opening Balance
> Assets:Checking$1000
> Equity:Opening
> 
> 2017/01/02 * Savings
> Assets:Savings  $100
> Assets:Checking
> 
> 2017/01/03 * Reconcile
> [Assets:Checking]  $0 = $900
> [Assets:Savings]   $0 = $100
> 
> 2017/01/04 * Savings
> Assets:Savings  $100
> Assets:Checking
> 
> 2017/01/05 * Reconcile
> [Assets:Checking]  $0 = $800
> [Assets:Savings]   $0 = $200
> 
> This works well. I read these reconcile transactions as "If $0 is added
> to
> the account, the balance should be $800".
> 
> If I instead change the last reconcile to "[Assets:Checking]  $0 = $700"
> (that is, make it $100 off) I get this error:
> 
> While parsing file "/home/mythmon/tmp/foo.ledger", line 18:
> While parsing posting:
>   [Assets:Checking]  $0 = $700
>   
> Error: Balance assertion off by $-100 (expected to see $800)
> 
> This shows me both the value I said it should be ($700), the value Ledger
> says it should be ($800), and the difference ($-100). All of these end up
> being very useful at different times.
> 
> I leave these in my ledger files. This gives me both a history of
> balances,
> and also a sort of test suite. If I change the organization of my files,
> change rules, or do other refactoring, these assertions give me
> confidence
> that I didn't break anything.
> 
> -Michael Cooper
> 
> 
> On Thu, Oct 19, 2017 at 3:43 PM John Lee  wrote:
> 
> > An example that currently has me puzzled: the first assert below passes,
> > and the second fails.  I'm interested both in learning why in this
> > particular case, and more important, learning how to use ledger to debug
> > problems like this in general.
> >
> > 2017-10-01 * Opening Balance
> > Assets:CurrentAccount   $1000
> > Equity:OpeningBalance
> >
> > 2017-10-02 * Savings
> > Assets:CurrentAccount:Savings   $100
> > Assets:CurrentAccount
> >
> > assert account("Assets:CurrentAccount").total == $1000
> >
> > 2017-10-03 * Savings
> > Assets:CurrentAccount:Savings   $100
> > Assets:CurrentAccount
> >
> > assert account("Assets:CurrentAccount").total == $1000
> >
> >
> > This surprises me because if I comment that last assert out and run
> > ledger b '^Assets:CurrentAccount$' I get $800, and ledger b
> > '^Assets:CurrentAccount:Savings$' prints $200 -- which adds to $1000,
> > and indeed ledger b '^Assets:CurrentAccount' prints $1000.  Bug??  How
> > can I see what ledger *thinks* the total is when it evaluates the
> > assert?
> >
> >
> > On Thu, 19 Oct 2017, at 23:13, John Lee wrote:
> > > Sometimes the assert expressions I'm adding to my ledger file fail and I
> > > don't immediately know why.  If I have something like this:
> > >
> > > assert account("Assets:CurrentAccount").total == $123.45
> > >
> > > and I'm wrong about the total, then I'll just be told that I'm wrong,
> > > 

Re: Debugging assertion failures / printing the value of a value expression

[Michael Cooper]

This isn't a direct answer to your question, but this is how I do these
kind of balance checks:

2017/01/01 * Opening Balance
Assets:Checking$1000
Equity:Opening

2017/01/02 * Savings
Assets:Savings  $100
Assets:Checking

2017/01/03 * Reconcile
[Assets:Checking]  $0 = $900
[Assets:Savings]   $0 = $100

2017/01/04 * Savings
Assets:Savings  $100
Assets:Checking

2017/01/05 * Reconcile
[Assets:Checking]  $0 = $800
[Assets:Savings]   $0 = $200

This works well. I read these reconcile transactions as "If $0 is added to
the account, the balance should be $800".

If I instead change the last reconcile to "[Assets:Checking]  $0 = $700"
(that is, make it $100 off) I get this error:

While parsing file "/home/mythmon/tmp/foo.ledger", line 18:
While parsing posting:
  [Assets:Checking]  $0 = $700
  
Error: Balance assertion off by $-100 (expected to see $800)

This shows me both the value I said it should be ($700), the value Ledger
says it should be ($800), and the difference ($-100). All of these end up
being very useful at different times.

I leave these in my ledger files. This gives me both a history of balances,
and also a sort of test suite. If I change the organization of my files,
change rules, or do other refactoring, these assertions give me confidence
that I didn't break anything.

-Michael Cooper


On Thu, Oct 19, 2017 at 3:43 PM John Lee  wrote:

> An example that currently has me puzzled: the first assert below passes,
> and the second fails.  I'm interested both in learning why in this
> particular case, and more important, learning how to use ledger to debug
> problems like this in general.
>
> 2017-10-01 * Opening Balance
> Assets:CurrentAccount   $1000
> Equity:OpeningBalance
>
> 2017-10-02 * Savings
> Assets:CurrentAccount:Savings   $100
> Assets:CurrentAccount
>
> assert account("Assets:CurrentAccount").total == $1000
>
> 2017-10-03 * Savings
> Assets:CurrentAccount:Savings   $100
> Assets:CurrentAccount
>
> assert account("Assets:CurrentAccount").total == $1000
>
>
> This surprises me because if I comment that last assert out and run
> ledger b '^Assets:CurrentAccount$' I get $800, and ledger b
> '^Assets:CurrentAccount:Savings$' prints $200 -- which adds to $1000,
> and indeed ledger b '^Assets:CurrentAccount' prints $1000.  Bug??  How
> can I see what ledger *thinks* the total is when it evaluates the
> assert?
>
>
> On Thu, 19 Oct 2017, at 23:13, John Lee wrote:
> > Sometimes the assert expressions I'm adding to my ledger file fail and I
> > don't immediately know why.  If I have something like this:
> >
> > assert account("Assets:CurrentAccount").total == $123.45
> >
> > and I'm wrong about the total, then I'll just be told that I'm wrong,
> > and not what the correct value is.  What's the easiest way to get ledger
> > to compute and print the correct value?
> >
> > Maybe there's a way to print out account values at a given point in a
> > ledger file?  In particular, maybe there's a way to print the values of
> > value expressions as the ledger file is evaluated by a command like
> > balance?
> >
> > I realize it's good to be able to see what the answer is before one
> > writes it down, but sometimes it just saves time to have the computer
> > work it out so you can see quickly what you did wrong.
>
> --
>
> ---
> You received this message because you are subscribed to the Google Groups
> "Ledger" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to ledger-cli+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"Ledger" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to ledger-cli+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: Debugging assertion failures / printing the value of a value expression

[John Lee]

An example that currently has me puzzled: the first assert below passes,
and the second fails.  I'm interested both in learning why in this
particular case, and more important, learning how to use ledger to debug
problems like this in general.

2017-10-01 * Opening Balance
Assets:CurrentAccount   $1000
Equity:OpeningBalance

2017-10-02 * Savings
Assets:CurrentAccount:Savings   $100
Assets:CurrentAccount

assert account("Assets:CurrentAccount").total == $1000

2017-10-03 * Savings
Assets:CurrentAccount:Savings   $100
Assets:CurrentAccount

assert account("Assets:CurrentAccount").total == $1000


This surprises me because if I comment that last assert out and run
ledger b '^Assets:CurrentAccount$' I get $800, and ledger b
'^Assets:CurrentAccount:Savings$' prints $200 -- which adds to $1000,
and indeed ledger b '^Assets:CurrentAccount' prints $1000.  Bug??  How
can I see what ledger *thinks* the total is when it evaluates the
assert?


On Thu, 19 Oct 2017, at 23:13, John Lee wrote:
> Sometimes the assert expressions I'm adding to my ledger file fail and I
> don't immediately know why.  If I have something like this:
> 
> assert account("Assets:CurrentAccount").total == $123.45
> 
> and I'm wrong about the total, then I'll just be told that I'm wrong,
> and not what the correct value is.  What's the easiest way to get ledger
> to compute and print the correct value?
> 
> Maybe there's a way to print out account values at a given point in a
> ledger file?  In particular, maybe there's a way to print the values of
> value expressions as the ledger file is evaluated by a command like
> balance?
> 
> I realize it's good to be able to see what the answer is before one
> writes it down, but sometimes it just saves time to have the computer
> work it out so you can see quickly what you did wrong.

-- 

--- 
You received this message because you are subscribed to the Google Groups 
"Ledger" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to ledger-cli+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.