The default behavior in Daml is to abort the transaction on any error and roll back all changes that have happened until then. However, this is not always appropriate. In some cases, it makes sense to recover from an error and continue the transaction instead of aborting it.
One option for doing that is to represent errors explicitly via
Option as shown in Data Types. This approach has
the advantage that it is very explicit about which operations are
allowed to fail without aborting the entire transaction. However, it
also has two major downsides. First, it can
be invasive for operations where aborting the transaction is often the
desired behavior, e.g., changing division to return
Either or an
Option to handle division by zero would be a very invasive change
and many callsites might not want to handle the error case explicitly.
Second, and more importantly, this approach does not allow rolling
back ledger actions that have happened before the point where failure is
detected; if a contract
got created before we hit the error, there is no way to undo that
except for aborting the entire transaction (which is what we were
trying to avoid in the first place).
By contrast, exceptions provide a way to handle certain types of errors in such a way that, on the one hand, most of the code that is allowed to fail can be written just like normal code, and, on the other hand, the programmer can clearly delimit which part of the current transaction should be rolled back on failure. All of that still happens within the same transaction and is thereby atomic contrary to handling the error outside of Daml.
Remember that you can load all the code for this section into a folder called
intro8 by running
daml new intro8 --template daml-intro-8
Our example for the use of exceptions will be a simple shop template. Users can order items by calling a choice and transfer money (in the form of an Iou issued by their bank) from their account to the owner in return.
First, we need to setup a template to represent the account of a user.
template Account with issuer : Party owner : Party amount : Decimal where signatory issuer, owner ensure amount > 0.0 key (issuer, owner) : (Party, Party) maintainer key._2 choice Transfer : () with newOwner : Party transferredAmount : Decimal controller owner, newOwner do create this with amount = amount - transferredAmount create Iou with issuer = issuer, owner = newOwner, amount = transferredAmount pure ()
Note that the template has an
ensure clause that ensures that the
amount is always positive so
Transfer cannot transfer more money
than is available.
The shop is represented as a template signed by the owner. It has a field to represent the bank accepted by the owner as well as a list of observers that can order items.
template Shop with owner : Party bank : Party observers : [Party] where signatory owner observer observers let price: Decimal = 100.0
The ordering process is then represented by a non-consuming choice on
this template which calls
Transfer and creates an
contract in return.
nonconsuming choice OrderItem : ContractId Order with shopper : Party controller shopper do exerciseByKey @Account (bank, shopper) (Transfer owner price) create Order with shopOwner = owner shopper = shopper
However, the shop owner has realized that often orders fail because
the account of their users is not topped up. They have a small trusted
userbase they know well so they decide that if the account is not
topped up, the shoppers can instead issue an Iou to the owner and pay
later. While it would be possible to check the conditions under which
Transfer will fail in
OrderItem this can be quite fragile: In
this example, the condition is relatively simple but in larger
projects replicating the conditions outside the choice and keeping the
two in sync can be challenging.
Exceptions allow us to handle this differently. Rather than
replicating the checks in
Transfer, we can instead catch the
exception thrown on failure. To do so we need to use a try-catch
try block defines the scope within which we want to
catch exceptions while the
catch clauses define which exceptions
we want to catch and how we want to handle them. In this case, we want
to catch the exception thrown by a failed
ensure clause. This
exception is defined in
PreconditionFailed. Putting it together our order process for
trusted users looks as follows:
nonconsuming choice OrderItemTrusted : ContractId Order with shopper : Party controller shopper do cid <- create Order with shopOwner = owner shopper = shopper try do exerciseByKey @Account (bank, shopper) (Transfer owner price) catch PreconditionFailed _ -> do create Iou with issuer = shopper owner = owner amount = price pure () pure cid
Let’s walk through this code. First, as mentioned, the shop owner is
the trusting kind, so he wants to start by creating the
matter what. Next, we try to charge the customer for the order. We
could, at this point, check their balance against the cost of the
order, but that would amount to duplicating the logic already present
Account. This logic is pretty simple in this case, but
duplicating invariants is a bad habit to get into. So, instead, we
just try to charge the account. If that succeeds, we just merrily
ignore the entire
catch clause; if that fails, however, we do not
want to destroy the Order contract we had already created. Instead, we
want to catch the error thrown by the
ensure clause of
Account (in this case, it is of type
try something else: create an
Iou contract to register the debt
and move on.
Note that if the
Iou creation still failed (unlikely with our
Iou here, but could happen in more complex
scenarios), because that one is not wrapped in a
try block, we
would revert to the default Daml behaviour and the
would be rolled back.
In addition to catching built-in exceptions like
PreconditionFailed, you can also define your own exception types
which can be caught and thrown. As an example, let’s consider a
variant of the
Transfer choice that only allows for transfers up
to a given limit. If the amount is higher than the limit, we throw an
We first have to define the exception and define a way to represent it as a string. In this case, our exception should store the amount that someone tried to transfer as well as the limit.
exception TransferLimitExceeded with limit : Decimal attempted : Decimal where message "Transfer of " <> show attempted <> " exceeds limit of " <> show limit
To throw our own exception, you can use
throwPure in other contexts.
choice TransferLimited : () with newOwner : Party transferredAmount : Decimal controller owner, newOwner do let limit = 50.0 when (transferredAmount > limit) $ throw TransferLimitExceeded with limit = limit attempted = transferredAmount create this with amount = amount - transferredAmount create Iou with issuer = issuer, owner = newOwner, amount = transferredAmount pure ()
Finally, we can adapt our choice to catch this exception as well:
nonconsuming choice OrderItemTrustedLimited : ContractId Order with shopper : Party controller shopper do try do exerciseByKey @Account (bank, shopper) (Transfer owner price) pure () catch PreconditionFailed _ -> do create Iou with issuer = shopper owner = owner amount = price pure () TransferLimitExceeded _ _ -> do create Iou with issuer = shopper owner = owner amount = price pure () create Order with shopOwner = owner shopper = shopper
For more information on exceptions, take a look at the language reference.
We have now seen how to develop safe models and how we can handle errors in those models in a robust and simple way. But the journey doesn’t stop there. In Work with Dependencies you will learn how to extend an already running application to enhance it with new features. In that context you’ll learn a bit more about the architecture of Daml, about dependencies, and about identifiers.