Never let your books land you in the pen.
Capone
is a library that provides double-entry bookkeeping (the foundation of
all modern accounting) for Django with the ability to link each recorded
transaction to zero or more other Django models as evidence for that
transaction.
In double-entry bookkeeping (DEB), all recordable events (purchases, sales,
equipment depreciation, bad debt markdowns, etc.) are tracked as "ledger
entries" or "transactions" in "ledgers". Each ledger entry is made up of one
or more "credit" and one or more "debit" entries. For the sake of this brief
example, you can think of credits as increasing the amount of money recorded in
a ledger and a debit decreasing it. With that assumption, the central idea behind
double-entry bookkeeping is that the sum of every ledger entry's debits must
equal the sum of its credits. capone
implements a double-entry bookkeeping
system by providing an API for checking that all created entries satisfy this
condition or rolling back the transaction if not.
In addition to this standard bookkeeping functionality, capone
also allows
any number of arbitrary objects to be attached, via generic foreign keys, to
a ledger entry as "evidence" for that transaction's having happened. For
instance, a transaction recording a bank deposit paying for several medical
tests at a time from an insurance company to your medical testing company could
be linked to the original Order
objects that recorded the test. capone
also provides an API for the efficient querying of ledger entries by evidence.
For more information on the concept of double-entry bookkeeping itself, we recommend the Wikipedia article: https://en.wikipedia.org/wiki/Double-entry_bookkeeping_system.
First, you must set up your working environment:
make setup
This will build a local virtualenv and all other requirements for local development.
The following commands are available for interacting with the app:
To start a shell instance so that you can interact with the app via the ORM:
make shell
Note: before any of the following instructions, you may have to run make develop
to set up a postgres database for this app.
First, activate a virtualenv so that your commands have access to the
environment built by make setup
:
From the repository root, run:
source .venv/bin/activate
Then you should be free to run
./manage.py makemigrations --settings=capone.tests.settings
or any other manage.py
command, even those in the Makefile.
To run individual tests, use the following:
./manage.py test --settings=capone.tests.settings capone.tests
Notice the --settings=capone.tests.settings
argument: because this repository
is a django sub-module, it wouldn't make sense for it to come with its own
default settings.py
file. Instead, it ships with one used to run its tests.
To use manage.py
, we have to pass an import path to the settings file
explicitly.
Let's introduce the models provided by capone
and how they relate to one
another.
Note that all objects in this library have created_at
and modified_at
fields that are auto_now_add
and auto_now
, respectively.
The models in this section are those that correspond most to well known accounting concepts, i.e. those involved in keeping accounts using the principles of double-entry bookkeeping. They model ledgers, journal entries, credits and debits, and any metadata one wishes to store with these objects.
A Ledger
is the top-most level of organization of information in double-entry
bookkeeping as well as the capone
app. Most ledgers have names familiar to
those with any knowledge of accounting, such as "revenue" or "accounts
receivable".
Ledgers
are synonymous with the accounting concept of an "account", so you
may see references to accounts in this documentation or elsewhere in the
accounting literature.
As a data structure, a Ledger
in this library is little more than a name,
description, and unique number: LedgerEntries
(see below) point to a Ledger
to represent their being "in" a Ledger
. Transactions
(see below also) that
are "between" two Ledgers
have a LedgerEntry
pointing to one Ledger
and
another LedgerEntry
pointing to the other Ledger
.
Ledger
also has the sometimes confusing field increased_by_debits
. All
Ledgers
are of one of two types: either debits increase the "value" of an
account or credits do. By convention, asset and expense accounts are of the
former type, while liabilities, equity, and revenue are of the latter: in
short, an increase to an "asset"-type account is a debit, and an increase to
a "liability" or "equity"-type account is a credit.
Here's a handy mnemonic for the two types of accounts: The accounting equation says (by definition) that:
assets == liabilities + owner equity
The terms on the right of the equals sign are increased by debits, and terms on the left of the equals sign are decreased by debits. We can therefore use the accounting equation to know whether to use debits or credits to model an increase in a ledger.
So because debits and credits mean different things in different types of accounts, we can have a transaction with an "equal and opposite" credit and debit pair of the same currency amount, but that still represents a net increase in the value of a company: a debit in Accounts Receivable and a credit in Revenue increases both accounts while satisfying the accounting equation.
Currently, field increased_by_debits
is not used by the code in capone
but
is provided as a convenience to users who might wish to incorporate this
information into an external report or calculation.
A Transaction
is a record of a discrete financial action, represented by
a collection of debits and credits whose sums equal one another. Practically
all models in capone
link to or through Transaction
: in a sense you could
say it's the main model provided by capone
. A Transaction
can sometimes be
referred to as a "journal entry".
The Transaction
model records debits and credits by linking to
LedgerEntries
, which include currency amounts of the proper sign, and those
LedgerEntries
themselves point to Ledger
. In other words, Transaction
and Ledger
are linked in a many-to-many fashion by going through
LedgerEntry
as a custom through model. The "proper sign" part is taken care
of by the credit
and debit
convenience methods (see examples below).
Transactions
should never be deleted. Instead, a new Transaction
with
debits and credits swapped should be created using
capone.api.actions.void_transaction
to negate the effect of the Transaction
you'd like to remove. The voids
field on the new Transaction
will
automatically be filled in with the old Transaction
you wish to remove. By
this method, you'll never have to delete data from your system as a part of
normal operation, which mimics one of the many benefits of traditional,
non-computerized double-entry bookkeeping.
Transaction
also has the following fields to provide metadata for each transaction:
created_by
: The user who created thisTransaction
.notes
: A free-form text field for adding to aTransaction
any information not expressed in the numerous metadata fields.posted_timestamp
: The time aTransaction
should be considered valid from.capone.api.actions.create_transaction
automatically deals with filling in this value with the current time. You can change this value to post-date or back-dateTransactions
becausecreated_at
will always represent the true object creation time.transaction_id
: A Universally Unique Identifier (UUID) for theTransaction
, useful for unambiguously referring to aTransaction
without using primary keys or other database internals.type
: A user-defined type for theTransaction
(see theTransactionType
model below).
A TransactionType
is a user-defined, human-readable "type" for
a Transaction
, useful for sorting, aggregating, or annotating Transactions
.
The default TransactionType
is MANUAL
, which is created automatically by
the library, but you can define others, say for bots or certain classes of
users.
Currently, TransactionType
is not used by the code in capone
but is
provided as a convenience to users who might wish to incorporate this
information into an external report or calculation.
LedgerEntries
represent single debit or credit entries in a single Ledger
.
LedgerEntries
are grouped together into Transactions
(see above) with the
constraint that the sum of all credit and debit LedgerEntries
for a given
Transaction
must equal zero.
LedgerEntries
have a field entry_id
, which is a UUID for unambiguously
referring to a single LedgerEntry
.
The models in this section deal with adding evidence to Transactions
and
searching over that evidence.
A TransactionRelatedObject
(TRO
) represents the "evidence" relationship that
makes the capone
library more useful. A TRO
links a Transaction
to an arbitrary
object in the larger app that this library is used in using a generic foreign
key. One TRO
links one Transaction
and one arbitrary object, so we make as
many TROs
as we want pieces of evidence. There are several convenience
methods in capone.api.queries
for efficiently querying over Transactions
based on evidence and evidence objects based on their Transactions
(see
examples below).
A LedgerBalance
is similar to a TRO
in that it allows linking ledger
objects with objects from the wider app that the library is used in via generic
foreign keys. The purpose of LedgerBalance
is to denormalize for more
efficient querying the current sum of debits and credits for an object in
a specific Ledger. Therefore, there is only one LedgerBalance
for each
(ledger, related_object)
tuple.
You should never have to manually create or edit a LedgerBalance
: doing so,
as well as keeping them up-to-date, is handled by capone
internals. For the
same reasons, deleting them is not necessary or a good idea.
The purpose of LedgerBalance
can best be demonstrated by considering the
deceptively simple query, "how many Orders (a non-capone
model we presumably
created in the app where we include capone
as a library) have an Accounts
Receivable balance greater than zero?" One would have to calculate the ledger
balance over literally the product of all ledgers and all non-capone
objects
in the database, and then filter them for all those with balances above zero,
to answer this question, which is obviously too expensive. By keeping track of
the per-Ledger
balance for each object used as evidence in a Transaction
,
we can much more easily make these queries with reasonable overhead.
Let's start by creating two common ledger types, "Accounts Receivable" and "Revenue", which usually have transactions between themselves:
>>> from capone.models import Ledger
>>> ar = Ledger.objects.create(name='Accounts Receivable', number=1, increased_by_debits=True)
<Ledger: Ledger Accounts Receivable>
>>> revenue = Ledger.objects.create(name='Revenue', number=2, increased_by_debits=True)
<Ledger: Ledger Revenue>
Both of these accounts are asset accounts, so they're both increased by debits.
Please consult the double-entry bookkeeping Wikipedia article or the
explanation for increased_by_debits
above for a more in-depth explanation of
the "accounting equation" and whether debits increase or decrease an account.
Also, note that the default convention in capone
is to store debits as
positive numbers and credits as negative numbers. This convention is common
but completely arbitrary. If you want to switch the convention around, you can
set DEBITS_ARE_NEGATIVE
to True
in your settings.py file. By default, that
constant doesn't need to be defined, and if it remains undefined, capone
will
interpret its value as False
.
Now let's create a fake Order, so that we have some evidence for these ledger entries, and a fake User, so we'll have someone to blame for these transactions:
>>> from capone.tests.factories import OrderFactory
>>> order = OrderFactory()
>>> from capone.tests.factories import UserFactory
>>> user = UserFactory()
We're now ready to create a simple transaction:
>>> from capone.api.actions import create_transaction
>>> from capone.api.actions import credit
>>> from capone.api.actions import debit
>>> from decimal import Decimal
>>> from capone.models import LedgerEntry
>>> txn = create_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(100)), ledger=revenue)])
>>> txn.summary()
{
u'entries': [
'LedgerEntry: $100.0000 in Accounts Receivable',
'LedgerEntry: $-100.0000 in Revenue',
],
u'related_objects': [
'TransactionRelatedObject: Order(id=1)',
]
}
Note that we use the helper functions credit
and debit
with positive
numbers to keep the signs consistent in our code. There should be no reason to
use negative numbers with capone
.
Note also that the value for the credit and debit is the same: $100. If we tried to create a transaction with mismatching amounts, we would get an error:
>>> create_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(101)), ledger=revenue)])
---------------------------------------------------------------------------
TransactionBalanceException Traceback (most recent call last)
[...]
TransactionBalanceException: Credits do not equal debits. Mis-match of -1.
So the consistency required of double-entry bookkeeping is automatically kept.
There are many other options for create_transaction
: see below or its
docstring for details.
capone
keeps track of the balance in each ledger for each evidence object in
a denormalized and efficient way. Let's use this behavior to get the balances
of our ledgers as well as the balances in each ledger for our order
object:
>>> from capone.api.queries import get_balances_for_object
>>> get_balances_for_object(order)
defaultdict(<function <lambda> at 0x7fd7ecfa96e0>, {<Ledger: Ledger Accounts Receivable>: Decimal('100.0000'), <Ledger: Ledger Revenue>: Decimal('-100.0000')})
>>> ar.get_balance()
Decimal('100.0000')
>>> revenue.get_balance()
Decimal('-100.0000')
We can also void that transaction, which enters a transaction with the same evidence but with all values of the opposite sign:
>>> from capone.api.actions import void_transaction
>>> void = void_transaction(txn, user)
<Transaction: Transaction 9cd85014-c588-43ff-9532-a6fc2429069e>
>>> void_transaction(txn, user)
---------------------------------------------------------------------------
UnvoidableTransactionException Traceback (most recent call last)
[...]
UnvoidableTransactionException: Cannot void the same Transaction #(e0842107-3a5b-4487-9b86-d1a5d7ab77b4) more than once.
>>> void.summary()
{u'entries': ['LedgerEntry: $-100.0000 in Accounts Receivable',
'LedgerEntry: $100.0000 in Revenue'],
u'related_objects': ['TransactionRelatedObject: Order(id=1)']}
>>> txn.voids
>>> void.voids
<Transaction: Transaction e0842107-3a5b-4487-9b86-d1a5d7ab77b4>
Note the new balances for evidence objects and Ledgers
:
>>> get_balances_for_object(order)
defaultdict(<function <lambda> at 0x7fd7ecfa9758>, {<Ledger: Ledger Accounts Receivable>: Decimal('0.0000'), <Ledger: Ledger Revenue>: Decimal('0.0000')})
>>> ar.get_balance()
Decimal('0.0000')
>>> revenue.get_balance()
Decimal('0.0000')
You can label a Transaction
using a foreign key to the TransactionType
to,
say, distinguish between manually made Transactions
and those made by a bot,
or between Transactions
that represent two different types of financial
transaction, such as "Reconciliation" and "Revenue Recognition".
By default, Transactions
are of a special, auto-generated "manual" type:
>>> txn.type
<TransactionType: Transaction Type Manual>
but you can create and assign TransactionTypes
when creating Transactions
:
>>> from capone.models import TransactionType
>>> new_type = TransactionType.objects.create(name='New type')
>>> txn = create_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(100)), ledger=revenue)], type=new_type)
>>> txn.type
<TransactionType: Transaction Type New type>
Transaction
has a summary
method to summarize the data on the many models
that can link to it:
>>> txn.summary()
{u'entries': ['LedgerEntry: $100.0000 in Accounts Receivable',
'LedgerEntry: $-100.0000 in Revenue'],
u'related_objects': ['TransactionRelatedObject: Order(id=1)']}
To get the balance for a Ledger
, use its get_balance
method:
>>> ar.get_balance()
Decimal('100.0000')
To efficiently get the balance of all transactions with a particular object as
evidence, use get_balances_for_objects
:
>>> get_balances_for_object(order)
defaultdict(<function <lambda> at 0x7fd7ecfa9230>, {<Ledger: Ledger Accounts Receivable>: Decimal('100.0000'), <Ledger: Ledger Revenue>: Decimal('-100.0000')})
Transactions
are validated before they are created, but if you need to do
this manually for some reason, use the validate_transaction
function, which
has the same prototype as create_transaction
:
>>> validate_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(100)), ledger=revenue)], type=new_type)
>>> validate_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(101)), ledger=revenue)], type=new_type)
---------------------------------------------------------------------------
TransactionBalanceException Traceback (most recent call last)
<ipython-input-64-07b6d139bb37> in <module>()
----> 1 validate_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(101)), ledger=revenue)], type=new_type)
/home/hunter/capone/capone/api/queries.pyc in validate_transaction(user, evidence, ledger_entries, notes, type, posted_timestamp)
67 if total != Decimal(0):
68 raise TransactionBalanceException(
---> 69 "Credits do not equal debits. Mis-match of %s." % total)
70
71 if not ledger_entries:
TransactionBalanceException: Credits do not equal debits. Mis-match of -1.
Along with the query possibilities from the Django ORM, capone
provides
Transaction.filter_by_related_objects
for finding Transactions
that are
related to certain models as evidence.
>>> Transaction.objects.count()
5
>>> Transaction.objects.filter_by_related_objects([order]).count()
5
>>> order2 = OrderFactory()
>>> create_transaction(user, evidence=[order2], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(100)), ledger=revenue)])
<Transaction: Transaction 68a4adb1-b898-493f-b5f3-4fe7132dd28d>
>>> Transaction.objects.filter_by_related_objects([order2]).count()
1
filter_by_related_objects
is defined on a custom QuerySet
provided for
Transaction
, so calls to it can be chained like ordinary QuerySet
function
calls:
>>> create_transaction(user, evidence=[order2], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(100)), ledger=revenue)])
<Transaction: Transaction 92049712-4982-4718-bc71-a405b0d762ac>
>>> Transaction.objects.filter_by_related_objects([order2]).count()
2
>>> Transaction.objects.filter_by_related_objects([order2]).filter(transaction_id='92049712-4982-4718-bc71-a405b0d762ac').count()
1
filter_by_related_objects
takes an optional match_type
argument, which is
of type MatchType(Enum)
that allows one to filter in different ways, namely
whether the matching transactions may have "any", "all", "none", or "exactly"
the evidence provided, determined by MatchTypes
ANY
, ALL
, NONE
, and
EXACT
, respectively.
For writing tests, the method
assert_transaction_in_ledgers_for_amounts_with_evidence
is provided for
convenience. As its name implies, it allows asserting the existence of exactly
one Transaction
with the ledger amounts, evidence, and other fields on Ledger
provided to the method.
>>> create_transaction(user, evidence=[order], ledger_entries=[LedgerEntry(amount=debit(Decimal(100)), ledger=ar), LedgerEntry(amount=credit(Decimal(100)), ledger=revenue)])
<Transaction: Transaction b3e73f1d-6b10-4597-b19b-84800839d5b3>
>>> with assert_raises(Transaction.DoesNotExist):
... assert_transaction_in_ledgers_for_amounts_with_evidence(ledger_amount_pairs=[(revenue.name, credit(Decimal(100))), (ar.name, debit(Decimal(100)))], evidence=[])
...
>>> assert_transaction_in_ledgers_for_amounts_with_evidence(ledger_amount_pairs=[(revenue.name, credit(Decimal(100))), (ar.name, debit(Decimal(100)))], evidence=[order])
>>> with assert_raises(Transaction.DoesNotExist):
... assert_transaction_in_ledgers_for_amounts_with_evidence(ledger_amount_pairs=[(revenue.name, credit(Decimal(100))), (ar.name, debit(Decimal(100)))], evidence=[order])
...
Traceback (most recent call last):
File "<console>", line 2, in <module>
File "/usr/lib/python2.7/unittest/case.py", line 116, in __exit__
"{0} not raised".format(exc_name))
AssertionError: DoesNotExist not raised
You can see
capone.tests.test_assert_transaction_in_ledgers_for_amounts_with_evidence
for
more examples!
Image courtesy Officer on Wikipedia. This work was created by a government unit (including state, county, and municipal government agencies) of the U.S. state of Florida. It is a public record that was not created by an agency which state law has allowed to claim copyright and is therefore in the public domain in the United States.