This fixes#1626.
The tokenizer eagerly classifies 4-digit integers as TOK_A_YEAR
tokens. In some contexts such as "every 1000 years", this causes
errors.
I think the tokenizer does not have enough information available to
distinguish between integers and years.
After this patch, the tokenizer will always classify integers as
TOK_INT tokens. The "has 4 digits" heuristic to determine if an
integer is a year is moved to the place where it's actually
needed (and it can be slightly more generic there, too).
This fixes issue #1127. In my understanding, PR #552 was meant to fix
this, but was incomplete.
Without this patch, automated transactions are invisible to
assertions.
This patch fixes this by adding a flag to the account to tell it that
there is a new posting, analogous to the behavior of finalize().
I dug up issue #1127 too late to find that this is the same solution
proposed by @tbm. Although I wrote this independently, credit goes to
Martin Michlmayr (@tbm).
`changed_value_posts::create_accounts()` reuses the `<Revalued>` account
from `display_filter`, but when clearing `changed_value_posts`
`create_accounts()` would be called before the account had been
recreated by `display_filter_posts`. This results in a segfault when
using the --group-by option.
I'm not sure if `display_filter_posts` has the same problem but I
reordered the calls there too for good measure.
* Re-initialize (to VOID) totals for the account and its ancestors on adding
postings. Otherwise the cache intended for use by recursive calls of C++
function total() in computing family (i.e. account hierarchy) totals is
incorrectly retained from one top-level call to the next, causing
inconsistent and broken behaviour.
* Re-initialize (to false) calculated and gathered. Otherwise we won't
e.g. recalculate stale totals for ancestor accounts (e.g. won't recalculate
Assets:Savings total if Assets:Savings changes via a posting).
Although the value expression total function is used by ledger itself in
computing totals, this bug would only appear on use of .total in user-supplied
value expressions computed *during parsing* of ledger files, rather than after
parsing (I believe ledger only ever calls it for internal purposes after
parsing is complete).
It is possible this bug also affected other functions than total (perhaps even
in circumstances other than analagous to that described in the preceding
paragraph). I have not checked that.
weeks - calculate start date for finding period using remainer 400/periodlength to reduce number of iterations (perhaps this ought to follow the same conventio as years months and quarters)
add sample period command tests
add add day period tests for forecasts and budgets
add week period tests for forecasts and budgets - these do not change
Bug 1159
Use interval start date (period from/since) to initialize first period.
This allows the offset of a period start to be specified
-p 'every 12 months from 2000-04-01'
will have periods
yyyy-04-01 to yyyy-03-31
....
If no from/since is specified for the period the first transaction date reported is used to initialize the period as before.
added test case regress/1159.test
