Canister REPL

ic-repl [--replica [local|ic|url] | --offline [--format [json|ascii|png]]] --config <toml config> [script file]

Commands

<command> := 
 | import <id> = <text> (as <text>)?                // bind canister URI to <id>, with optional did file
 | export <text>                                    // export current environment variables
 | load <text>                                      // load and run a script file
 | config <text>                                    // set config in TOML format
 | let <id> = <exp>                                 // bind <exp> to a variable <id>
 | <exp>                                            // show the value of <exp>
 | assert <exp> <binop> <exp>                       // assertion
 | identity <id> (<text> | record { slot_index = <nat>; key_id = <text> })?   // switch to identity <id>, with optional pem file or HSM config
 | function <id> ( <id>,* ) { <command>;* }         // define a function
 | if <exp> { <command>;* } else { <command>;* }    // conditional branch
 | while <exp> { <command>;* }                      // while loop
<exp> := 
 | <candid val>                                     // any candid value
 | <var> <transformer>*                             // variable with optional transformers
 | fail <exp>                                       // convert error message as text
 | call (as <name>)? <name> . <name> (( <exp>,* ))? // call a canister method, and store the result as a single value
 | encode (<name> . <name>)? (( <exp>,* ))?         // encode candid arguments as a blob value. canister.__init_args represents init args
 | decode (as <name> . <name>)? <exp>               // decode blob as candid values
 | <id> ( <exp>,* )                                 // function application
<var> := 
 | <id>                  // variable name 
 | _                     // previous eval of exp is bind to `_`
<transformer> :=
 | ?                     // select opt value
 | . <name>              // select field name from record or variant value
 | [ <exp> ]             // select index from vec, text, record, or variant value
 | . <id> ( <exp>,* )    // transform (map, filter, fold) a collection value
<binop> := 
 | ==                    // structural equality
 | ~=                    // equal under candid subtyping; for text value, we check if the right side is contained in the left side
 | !=                    // not equal

Functions

Similar to most shell languages, functions in ic-repl is dynamically scoped and untyped.

We also provide some built-in functions:

account(principal): convert principal to account id.
neuron_account(principal, nonce): convert (principal, nonce) to account in the governance canister.
file(path): load external file as a blob value.
gzip(blob): gzip a blob value.
stringify(exp1, exp2, exp3, ...): convert all expressions to string and concat. Only supports primitive types.
output(path, content): append text content to file path.
wasm_profiling(path)/wasm_profiling(path, record { trace_only_funcs = <vec text>; start_page = <nat>; page_limit = <nat> }): load Wasm module, instrument the code and store as a blob value. Calling profiled canister binds the cost to variable __cost_{id} or __cost__. The second argument is optional, and all fields in the record are also optional. If provided, trace_only_funcs will only count and trace the provided set of functions; start_page writes the logs to a preallocated pages in stable memory; page_limit specifies the number of the preallocated pages, default to 4096 if omitted. See ic-wasm's doc for more details.
flamegraph(canister_id, title, filename): generate flamegraph for the last update call to canister_id, with title and write to {filename}.svg. The cost of the update call is returned.
concat(e1, e2): concatenate two vec/record/text together.
add/sub/mul/div(e1, e2): addition/subtraction/multiplication/division of two integers/floats. If one of the arguments is float32/float64, the result is float64; otherwise, the result is integer. You can use type annotation to get the integer part of the float number. For example div((mul(div(1, 3.0), 1000) : nat), 100.0) returns 3.33.
lt/lte/gt/gte(e1, e2): check if integer/float e1 is less than/less than or equal to/greater than/greater than or equal to e2.
eq/neq(e1, e2): check if e1 and e2 are equal or not. e1 and e2 must have the same type.
and/or(e1, e2)/not(e): logical and/or/not.
exist(e): check if e can be evaluated without errors. This is useful to check the existence of data, e.g., exist(res[10]).
ite(cond, e1, e2): expression version of conditional branch. For example, ite(exist(res.ok), "success", "error").

The following functions are only available in non-offline mode:

read_state([effective_id,] prefix, id, paths, ...): fetch the state tree path of <prefix>/<id>/<paths>. Some useful examples,
- candid metadata: read_state("canister", principal "canister_id", "metadata/candid:service")
- canister controllers: read_state("canister", principal "canister_id", "controllers")
- list all subnet ids: read_state("subnet")
- subnet metrics: read_state("subnet", principal "subnet_id", "metrics")
- list subnet nodes: read_state("subnet", principal "subnet_id", "node")
- node public key: read_state("subnet", principal "subnet_id", "node", principal "node_id", "public_key")
send(blob): send signed JSON messages generated from offline mode. The function can take a single message or an array of messages. Most likely use is send(file("messages.json")). The return result is the return results of all calls. Alternatively, you can use ic-repl -s messages.json -r ic.

Object methods

For vec, record or text value, we provide some built-in methods for value transformation:

v.map(func): transform each item v[i] with func(v[i]).
v.filter(func): filter out item v[i] if func(v[i]) returns false or has an error.
v.fold(init, func): combine all items in v by repeatedly applying func(...func(func(init, v[0]), v[1])..., v[n-1]).
v.size(): count the size of v.

For record value, v[i] is represented as record { key; value } sorted by field id.

For text value, v[i] is represented as a text value containing a single character.

Type casting

Type annotations in ic-repl is more permissible (not following the subtyping rules) than the Candid library to allow piping results from different canister calls.

("text" : blob) becomes blob "text" and vice versa. Converting blob to text can get an error if the blob is not utf8 compatible.
(service "aaaaa-aa" : principal) becomes principal "aaaaa-aa". You can convert among service, principal and func.
((((1.99 : nat8) : int) : float32) : nat32) becomes (1 : nat32). When converting from float to integer, we only return the integer part of the float.
Type annotations for record, variant is left unimplemented. With candid interface embedded in the canister metadata, annotating composite types is almost never needed.

Examples

test.sh

#!/usr/bin/ic-repl -r ic
// assume we already installed the greet canister
import greet = "rrkah-fqaaa-aaaaa-aaaaq-cai";
call greet.greet("test");
let result = _;
assert _ == "Hello, test!";
identity alice;
call "rrkah-fqaaa-aaaaa-aaaaq-cai".greet("test");
assert _ == result;

nns.sh

#!/usr/bin/ic-repl -r ic
// nns and ledger canisters are auto-imported if connected to the mainnet
call nns.get_pending_proposals()
identity private "./private.pem";
call ledger.account_balance(record { account = account(private) });

function transfer(to, amount, memo) {
  call ledger.transfer(
    record {
      to = to;
      fee = record { e8s = 10_000 };
      memo = memo;
      from_subaccount = null;
      created_at_time = null;
      amount = record { e8s = amount };
    },
  );
};
function stake(amount, memo) {
  let _ = transfer(neuron_account(private, memo), amount, memo);
  call nns.claim_or_refresh_neuron_from_account(
    record { controller = opt private; memo = memo }
  );
  _.result?.NeuronId
};
let neuron_id = stake(100_000_000, 42);

install.sh

#!/usr/bin/ic-repl
function deploy(wasm) {
  let id = call ic.provisional_create_canister_with_cycles(record { settings = null; amount = null });
  call ic.install_code(
    record {
      arg = encode wasm.__init_args();
      wasm_module = wasm;
      mode = variant { install };
      canister_id = id.canister_id;
    },
  );
  id
};

identity alice;
let id = deploy(file("greet.wasm"));
let status = call ic.canister_status(id);
assert status.settings ~= record { controllers = vec { alice } };
assert status.module_hash? == blob "...";
let canister = id.canister_id;
call canister.greet("test");

wallet.sh

#!/usr/bin/ic-repl
import wallet = "${WALLET_ID:-rwlgt-iiaaa-aaaaa-aaaaa-cai}" as "wallet.did";
identity default "~/.config/dfx/identity/default/identity.pem";
call wallet.wallet_create_canister(
  record {
    cycles = ${CYCLE:-1_000_000};
    settings = record {
      controllers = null;
      freezing_threshold = null;
      memory_allocation = null;
      compute_allocation = null;
    };
  },
);
let id = _.Ok.canister_id;
call as wallet ic.install_code(
  record {
    arg = encode ();
    wasm_module = file("${WASM_FILE}");
    mode = variant { install };
    canister_id = id;
  },
);
call id.greet("test");

profiling.sh

#!/usr/bin/ic-repl
import "install.sh";

let file = "result.md";
output(file, "# profiling result\n\n");
output(file, "|generate|get|put|\n|--:|--:|--:|\n");

let cid = deploy(gzip(wasm_profiling("hashmap.wasm")));
call cid.__toggle_tracing();   // Disable flamegraph tracing
call cid.generate(50000);
output(file, stringify(__cost__, "|"));

call cid.__toggle_tracing();   // Enable flamegraph tracing
call cid.batch_get(50);
flamegraph(cid, "hashmap.get(50)", "get");
output(file, stringify("[", __cost__, "](get.svg)|"));

let put = call cid.batch_put(50);
flamegraph(cid, "hashmap.put(50)", "put.svg");
output(file, stringify("[", __cost_put, "](put.svg)|\n"));

recursion.sh

function fib(n) {
  let _ = ite(lt(n, 2), 1, add(fib(sub(n, 1)), fib(sub(n, 2))))
};
function fib2(n) {
  let a = 1;
  let b = 1;
  while gt(n, 0) {
      let b = add(a, b);
      let a = sub(b, a);
      let n = sub(n, 1);
  };
  let _ = a;
};
function fib3(n) {
  if lt(n, 2) {
      let _ = 1;
  } else {
      let _ = add(fib3(sub(n, 1)), fib3(sub(n, 2)));
  }
};
assert fib(10) == 89;
assert fib2(10) == 89;
assert fib3(10) == 89;

Relative paths

Several commands and functions are taking arguments from the file system. We have different definitions for relative paths, depending on whether you are reading or writing the file.

For reading files, e.g., import, load, identity, file, wasm_profiling, relative paths are based on where the current script is located;
For writing files, e.g., export, output, flamegraph, relative paths are based on the current directory when the script is run.

The rationale for the difference is that we can have an easier time to control where the output files are located, as scripts can spread out in different directories.

Derived forms

call as proxy_canister target_canister.method(args) is a shorthand for

let _ = call proxy_canister.wallet_call(
  record {
    args = encode target_canister.method(args);
    cycles = 0;
    method_name = "method";
    canister = principal "target_canister";
  }
);
decode as target_canister.method _.Ok.return

Canister init args types

When calling ic.install_code, you may need to provide a Candid message for initializing the canister. To help with encoding the message, you can use get the init args types from the Wasm module custom section:

let wasm = file("a.wasm");
encode wasm.__init_args(...)

If the Wasm module doesn't contain the init arg types, you can import the full did file as a workaround:

import init = "2vxsx-fae" as "did_file_with_init_args.did";
encode init.__init_args(...)

Contributing

Please follow the guidelines in the CONTRIBUTING.md document.

Issues

Autocompletion within Candid value
Robust support for ~=, requires inferring principal types
Loop detection for load
Assert upgrade correctness

dfinity / ic-repl