6. Wallet Manual

The GNU Taler wallet allows customers to withdraw and spend digital cash.

6.1. WebExtension Wallet

6.1.1. Building from source

$ git clone https://git.taler.net/wallet-core.git
$ cd wallet-core
$ ./configure
$ make webex-stable
# Packaged extension now available as:
# dist/taler-wallet-$VERSION.zip

6.2. Android Wallet

Please see Building apps from source in the Developer’s Manual.

6.3. iOS Wallet

Please see Building Taler Wallet for iOS from source in the Developer’s Manual.

6.4. Command-line Wallet

This section describes how to use the GNU Taler wallet command line interface (CLI).

The the wallet CLI is targeted at developers and operators, but not meant to be used by customers. It exposes all functionality that the more user-friendly interfaces (Android app, browser extension) offer. However, it provides more diagnostics and advanced features as well.

6.4.1. Building from source

The easiest way to install the wallet is via NPM. Note that a recent version of Node.JS (>=12.20.1) is required.

We recommend to install the wallet package on a per-user basis, thus setting $INSTALL_PREFIX to a directory in $HOME.

$ git clone https://git.taler.net/wallet-core.git
$ cd wallet-core
$ ./bootstrap
$ ./configure --prefix=$INSTALL_PREFIX
$ make && make install

The wallet command-line interface should then be available as taler-wallet-cli under $INSTALL_PREFIX/bin.

6.4.2. Installation via NPM

The wallet can also obtained via NPM, the Node Package Manager.

To install the wallet as a global package, run:

$ npm install -g taler-wallet
# check if installation was successful
$ taler-wallet-cli --version

To install the wallet only for your user, run:

$ npm install -g --prefix=$HOME/local taler-wallet
# check if installation was successful
$ taler-wallet-cli --version
# If this fails, make sure that $HOME/local/bin is in your $PATH

To use the wallet as a library in your own project, run:

$ npm install taler-wallet

6.4.3. Getting Help

The wallet CLI comes with built-in help. Invoke the wallet CLI (or any subcommand) with the --help flag to get help:

$ taler-wallet-cli --help
Usage: taler-wallet-cli COMMAND

Command line interface for the GNU Taler wallet.

Options:
  -h, --help             Show this message and exit.
  --wallet-db=VALUE      location of the wallet database file
  --timetravel=VALUE     modify system time by given offset in microseconds
  --inhibit=VALUE        Inhibit running certain operations, useful for debugging and testing.
  --no-throttle          Don't do any request throttling.
  -v, --version
  -V, --verbose          Enable verbose output.

Commands:
  advanced               Subcommands for advanced operations (only use if you know what you're doing!).
  api                    Call the wallet-core API directly.
  backup                 Subcommands for backups
  balance                Show wallet balance.
  deposit                Subcommands for depositing money to payto:// accounts
  exchanges              Manage exchanges.
  handle-uri             Handle a taler:// URI.
  pending                Show pending operations.
  run-pending            Run pending operations.
  run-until-done         Run until no more work is left.
  testing                Subcommands for testing GNU Taler deployments.
  transactions           Show transactions.

6.4.4. Completing operations

Note that the CLI does not run as a background daemon. When starting operations that don’t immediately finish, the wallet needs to be run explicitly to finish any pending tasks:

# Do one attempt to finish all pending operations
$ taler-wallet-cli run-pending

# Run until all work is done
$ taler-wallet-cli run-until-done

6.4.5. Resetting the wallet

The wallet can be reset by deleting its database file. By default, the database file is $HOME/.talerwalletdb.json.

6.4.6. Handling taler:// URIs

Many interactions with the Taler wallet happen by scanning QR codes or special headers on Websites. To emulate this with the command line interface, run the following command:

$ taler-wallet-cli handle-uri $URI

6.4.7. Manual withdrawing

$ taler-wallet-cli advanced withdraw-manually \
    --exchange https://exchange.eurint.taler.net/ \
    --amount EUR:5

6.4.8. P2P push payments

The following code generates a P2P push transaction over 1 CHF with an expiration time of 30 days (assuming the wallet has a sufficient balance):

$ taler-wallet-cli p2p initiate-push-debit \
    --purse-expiration="30 d" \
    --summary="The summary" \
    CHF:1

The final URL can then be found in the transaction list:

$ taler-wallet-cli transactions

6.4.9. Background wallet

A wallet can be launched in the background:

$ taler-wallet-cli advanced serve &

You can then run various Taler operations faster against this one persistent instance:

$ taler-wallet-cli --wallet-connection=wallet-core.sock ...

Here ... needs to be changed to the commando to run. Make sure to run

$ taler-wallet-cli --wallet-connection=wallet-core.sock \
    run-until-done

to wait for pending transactions to complete.

6.4.10. Testing an exchange deployment

The following series of commands can be used to check that an exchange deployment is functional:

# This will now output a payto URI that money needs to be sent to in order to allow withdrawal
# of taler coins
$ taler-wallet-cli advanced withdraw-manually --exchange $EXCHANGE_URL --amount EUR:10.50

# Show the status of the manual withdrawal operation
$ taler-wallet-cli transactions

# Once the transfer has been made, try completing the withdrawal
$ taler-wallet-cli run-pending

# Check status of transactions and show balance
$ taler-wallet-cli transactions
$ taler-wallet-cli balance

# Now, directly deposit coins with the exchange into a target account
# (Usually, a payment is made via a merchant.  The wallet provides
# this functionality for testing.)
$ taler-wallet-cli deposit create EUR:5 payto://iban/$IBAN

# Check if transaction was successful.
# (If not, fix issue with exchange and run "run-pending" command again)
$ taler-wallet-cli transactions

# The wallet can also track if the exchange wired the money to the merchant account.
# The "deposit group id" can be found in the output of the transactions list.
$ taler-wallet-cli deposit track $DEPOSIT_GROUP_ID

6.5. Paying for an order

Note

This section is in dire need for some editing…

This section explains how to pay for an order in a scenario where the fiat bank is simulated. The simulation takes place by crafting ad-hoc XML files as if the bank would have issued them. Such XML files carry information about incoming payments to the regional currency master bank account. Finally, the XML files are passed to LibEuFin nexus via a convenient CLI method. The responsible script for such simulation is withdraw.sh.

Run ./withdraw.sh without any arguments. Assuming that you ran the command as the test-user, after the execution, 5 units of the regional currency should be found in the CLI wallet owned by test-user.

Check it with:

$ taler-wallet-cli balance

If so, call the wallet in the following way to finally pay for the order just created:

$ taler-wallet-cli handle-uri "$TALER_PAY_URI"

Note

Reset the state before going to production, as it impacts the way nexus asks records to the bank. In particular, delete: any database and the files config/user.conf and config/internal.conf, and finally run ./main.sh again.

6.6. APIs and Data Formats

6.6.1. Envelope Format

All API responses and notifications are returned in the following envelope:

type WalletResponseEnvelope =
 | WalletSuccess
 | WalletError
 | WalletNotification
export interface WalletSuccess {
  type: "response";
  operation: string;
  // ID to correlate success response to request
  id: string;
  // Result type depends on operation
  result: unknown;
}
export interface WalletError {
  type: "error";
  operation: string;
  // ID to correlate error response to request
  id: string;
  error: WalletErrorInfo;
}
export interface WalletSuccess {
  type: "notification";

  // actual type is WalletNotification,
  // to be documented here
  payload: any;
}
export interface WalletErrorInfo {
  // Numeric error code defined defined in the
  // GANA gnu-taler-error-codes registry.
  talerErrorCode: number;

  // English description of the error code.
  talerErrorHint: string;

  // English diagnostic message that can give details
  // for the instance of the error.
  message: string;

  // Error details, type depends
  // on talerErrorCode
  details: unknown;
}

6.6.2. Withdrawal

A typical API sequence for bank-integrated withdrawals can for example look like this:

  1. "getWithdrawalDetailsForUri" returns an amount and default exchange

  2. "getWithdrawalDetailsForAmount" returns fee information and that ToS are not accepted

    1. "getExchangeTos" are shown to the user and return currentEtag

    2. "setExchangeTosAccepted" called with currentEtag after user accepted

  3. "acceptWithdrawal" after the user confirmed withdrawal with associated fees

A typical API sequence for manual withdrawals can for example look like this:

  1. "listExchanges" shows a list of exchanges to the user who picks one and an amount

  2. "getWithdrawalDetailsForAmount" returns fee information and that ToS are not accepted

    1. "getExchangeTos" are shown to the user and return currentEtag

    2. "setExchangeTosAccepted" called with currentEtag after user accepted

  3. "acceptManualWithdrawal" after the user confirmed withdrawal with associated fees

6.7. Integration Tests

6.7.1. Integration Test Example

Integration tests can be done with the low-level wallet commands. To select which coins and denominations to use, the wallet can dump the coins in an easy-to-process format (CoinDumpJson).

The database file for the wallet can be selected with the --wallet-db option. This option must be passed to the taler-wallet-cli command and not the subcommands. If the database file doesn’t exist, it will be created.

The following example does a simple withdrawal recoup:

# Withdraw digital cash
$ taler-wallet-cli --wallet-db=mydb.json testing withdraw \
    -b https://bank.int.taler.net/ \
    -e https://exchange.int.taler.net/ \
    -a INTKUDOS:10

$ coins=$(taler-wallet-cli --wallet-db=mydb.json advanced dump-coins)

# Find coin we want to revoke
$ rc=$(echo "$coins" | \
       jq -r '[.coins[] | select((.denom_value == "INTKUDOS:5"))][0] | .coin_pub')

# Find the denom
$ rd=$(echo "$coins" | \
       jq -r '[.coins[] | select((.denom_value == "INTKUDOS:5"))][0] | .denom_pub_hash')

# Find all other coins, which will be suspended
$ susp=$(echo "$coins" | \
         jq --arg rc "$rc" '[.coins[] | select(.coin_pub != $rc) | .coin_pub]')

# The exchange revokes the denom
$ taler-exchange-keyup -r $rd
$ taler-deployment-restart

# Now we suspend the other coins, so later we will pay with the recouped coin
$ taler-wallet-cli --wallet-db=mydb.json advanced suspend-coins "$susp"

# Update exchange /keys so recoup gets scheduled
$ taler-wallet-cli --wallet-db=mydb.json exchanges update -f https://exchange.int.taler.net/

# Block until scheduled operations are done
$ taler-wallet-cli --wallet-db=mydb.json run-until-done

# Now we buy something, only the coins resulting from recouped will be
# used, as other ones are suspended
$ taler-wallet-cli --wallet-db=mydb.json testing test-pay \
    -m https://backend.int.taler.net/ \
    -k sandbox \
    -a "INTKUDOS:1" \
    -s "foo"
$ taler-wallet-cli --wallet-db=mydb.json run-until-done

To test refreshing, force a refresh:

$ taler-wallet-cli --wallet-db=mydb.json advanced force-refresh "$coin_pub"

To test zombie coins, use the timetravel option. It must be passed to the top-level command and not the subcommand:

# Update exchange /keys with time travel, value in microseconds
$ taler-wallet-cli --timetravel=1000000 --wallet-db=mydb.json \
    exchanges update -f https://exchange.int.taler.net/

6.7.2. Test Cases

Things we already have tests for:

  • Can the wallet recoup coins and spend them? [link]

Things we still need tests for:

  • Does the wallet do retries correctly when the exchange is not reachable? Or when the merchant is not reachable? Or the bank? This can be tested by temporarily killing those services.

  • How does the wallet deal with processing the same taler://(pay|withdraw) URI twice?

  • Test refunds

  • Test for session-based payments

  • Test case for auto-refunds (scenario where the vending machine finds out that its motor is broken, so it automatically gives a refund)

  • Does the wallet report “insufficient balance” correctly (as opposed to, say, crashing)?

  • Perf tests: How does the wallet handle withdrawing a LOT of coins?

  • Are the transaction history and pending operations reported correctly?

Tests for things the wallet doesn’t handle correctly yet:

  • What happens if the wallet double-spends a coin? (Easy to test by copying the wallet DB before spending and then running a spend again with the old DB).

  • What happens when a reserve is changed between accepting withdrawal and actually withdrawing coins? (This is harder to test. Might not be possible with the current CLI. The idea would be be to have some --inhibit=withdraw flag that tells the wallet to not actually withdraw, so we can change the reserve state and then resume the wallet.)

  • What happens if the exchange suddenly has a completely new list of denominations on offer?

  • What happens if the exchange changes its master public key? The wallet should handle this gracefully even if we have coins with that exchange, provided that the old denominations can be recouped. (That one is pretty difficult!)

  • Does the wallet handle payment aborts correctly?

There are test cases that require us to modify the communication between the wallet and exchange.

  • What does the wallet do when the exchange/merchant announce an incompatible protocol version?

  • What happens if some signature made by the exchange/merchant is garbage?

  • What if the exchange reports a double-spend and the proof it gives us is invalid?

6.7.3. Integration Test and Fault Injection Framework

This section describes the current approach to integration testing in the wallet.

It’s all based on a TypeScript harness process, which itself implements the fault injection proxy (async and in-process)!

The new approach consists of the following parts:

1. A strongly typed, convenient helper library to easily set up and run arbitrary Taler deployments and run test cases. These components plug together as easily as lego bricks, even with multiple exchanges/merchants/banks/etc. Logs and clean shutdown (even on SIGINT or errors) are handled properly. (Support for auditors is still pending but needed to fully test the wallet.)

This is how a simple withdrawal and payment test case looks like: https://git.taler.net/wallet-core.git/tree/packages/taler-integrationtests/src/test-payment.ts

(What’s particularly nice is that all our docs contain TypeScript definitions for all API request bodies. So just copying them into the test harness gives us auto-completion and compile-time checks to avoid typos. The wallet’s JSON validation machinery is also re-used.)

2. A fault injection proxy that can be plugged between the services and/or the wallet. It runs alongside the test harness, and can thus can use arbitrary custom logic. There’s no dependency for it other than built-in Node.JS libraries. Simple fault injections are just as easy to set up as with the twister.

The following test case (a) logs all requests and responses to the test harness stdout and (b) at a certain point, starts dropping the next 10 requests to the exchange (testing the wallet’s retry logic):

https://git.taler.net/wallet-core.git/tree/packages/taler-integrationtests/src/test-payment-fault.ts#n165

3. All util functionality from JS wallet-core, such as the Taler crypto, amount/date/etc. handling and JSON parsing/validation (the wallet is now more modular and easier to use as a library) can be used in the integration tests, even if a different wallet (Kotlin, whatever) is tested via the CLI.

4. A bunch of test cases that use (1)-(3). These are significantly more readable and hackable than other test approaches we had, while allowing for more complex scenarios. There are still way too few tests though!

5. A test runner (written in bash) that runs test cases based on a glob pattern and reports the results.

Injecting a fault is as easy as:

// Set up test case
[...]

exchangeProxy.addFault({
  beforeResponse(ctx: FaultInjectionResponseContext) {
     if (cond1) { // Drop some responses
       ctx.dropResponse = true;
       return;
     } else if (cond2) { // modify some others
       ctx.responseBody = Buffer.from(`{"oops": true}`, "utf-8");
       return;
     }
     // Other things that can be modified:
     // - drop/modify the request, not just the response
     // - modify headers
     // - modify status codes
  }
});

await doSomethingWithTheWallet();

exchangeProxy.clearFault();

await doMoreWithTheWallet();

To make the configuration easy, an ExchangeService (or MerchantService, BankService etc.) can be wrapped in a FaultInjectedExchangeService, which implements the ExchangeServiceInterface:

// create exchange and two merchants
const exchange = await setupExchange(...);
const merchant1 = ...;
const merchant2 = ...;

// Add exchange to merchant-accepted exchanges.
// This will adjust the config.
merchant1.addExchange(exchange);

// Wrap exchange in fault injection proxy
const faultInjectedExchange: ExchangeServiceInterface
  = new FaultInjectedExchangeService(t, exchange1, 8085);

// Merchant 2 talks to the exchange over fault injection,
// and thus must use the "twisted" base URL.
merchant2.addExchange(faultInjectedExchange);

The package for the integration tests is here:

https://git.taler.net/wallet-core.git/tree/packages/taler-integrationtests

The shortcut to run all integration tests is

./bootstrap && ./configure --prefix=... \
   && make install integrationtests

(From the root of the whole repo. If you’re developing tests, it’s way faster to just run “make compile install” once and then use “./testrunner” from the taler-integrationtests package.)