much of this chapter was populated with ngn's help.
One thing that array languages used to be (and still are) notorious for are their terse error messages. This is often a hill to climb for most programmers. APL, J and K4 try to work around this problem with their debugging tools, but the common thing that they all share is that errors are not very descriptive. Array language expect you to use your knowledge of the symbols to guide the debugging process.
K does not have complex debugging tools available, and primarily favours print debugging. The types of errors that can happen are small in number, generally provide the position of the error, and errors take up very less space on screen. Hence, K debugging is usually simpler than one expects. If you like long, detailed explanations of what you need to do, then K may not be for you.
This page will disambiguate all the possible errors that ngn/k can throw, and what can cause them.
A K error will generally have a) an error name and b) A message prefixed with a quote and b) a few lines pointing to the code where the error happened. Errors will be disambiguated first on error name, then the situation it may appear in.
A parse error means that the K interpreter has found malformed code. This is also called a syntax error.
Examples:
/ mismatched bracket:
({
/ unclosed string:
"fsdfsad
/ bad parameter lists:
{[x;1]1}
/ bad hex literals:
0xfgh
/ odd number of digits in hex:
0x123
/ nonexistent io verbs:
9:`
A rank error occurs when a primitive is called with arguments it does not have an overload for. This can be due to mismatched argument types, or a mismatched number of arguments.
Example:
/ + takes only 1 or 2 args:
+[1;2;3;4]
When the type of a value is incompatible with a primitive, a type error is thrown.
Examples:
1+`a
!"a"
&`symbol
A domain error occurs when the argument types for a primitive match, their lengths are correct, but the values of the arguments are not supported (they are not in the right domain).
Examples:
(0;1 2)!3 / arrays cannot be keys
&-1 / cannot repeat by negative amount
1?-1 / negative range does not work
1 2_() / empty array cannot be dropped from
-1 2_(1 2;3 4) / negative drop values do not work on a 2d array
An index error triggers when you assign outside the bound of an array. This can be done by normal assignment, or with the help of amend.
Example:
a:1 2 3;a[-7]:0 / assigning to an index that does not exist
@[1 2 3;0 5;:;8 9] / amending an index that does not exist
A length error occurs when the lengths of two arguments do not conform to each other in an each (') or a conforming operation.
Example: 1 2 3 + 1 2
A value error occurs on variable names that are not declared yet.
Example: a+5 will error if a is not defined.
NYI means not yet implemented. It triggers on overloads which are in ngn/k's design, but are not implemented yet.
'nyi errors change from time to time, and if you do get one, you have to find a workaround for
the primitive you are trying to use.
A stack overflow happens when a function recursively calls itself too much.
Example:
{o}[]
{o}
{o[]}[]
'stack
{o[]}
^
{o[]}
^
{o[]}
^
{o[]}
^
..
An IO error happens when there is a problem during the usage of an I/O verb. This can be:
- A file permissions problem
- file access problem
- problem with the path given to the verb
ngn/k first compiles your code to bytecode. This indicates a problem in the bytecode compilation phase.
All functions in ngn/k K have the following limits:
- 8 local variables (including parameters)
- max bytecode size per function - 256 bytes
If a function exceeds these limits, you will get this error.
Example:
{[a;b;c;d;e;f;g;h;i]a} / i is the 9th variable
Every K program must end with a trailing empty line. If you get this error often, you may want to ask your editor to insert one. Most popular editors have a way to do it.
Out of memory. Your program seems to be gobbling up large amounts of memory. This often happens in conditional iteration where values tend to get really big.
oom is a non recoverable error i.e. K will not point to the line where everything went wrong, so you'll need to check for the location yourself.
Print debugging is the main method of debugging in K. The general method of printing ( `0:) is quite clunky in this regard, so instead you should use \ (trace, note the leading space),
which autoformats and prints the K value given to it, and then returns it. Trace is an adverb, and hence will be executed like an adverb, right to left.
You can also throw custom errors using `err, which will throw an error with the name of the string given, along with the line the error was thrown on.
`err@"FsdfsdfA"
'FsdfsdfA
`err@"FsdfsdfA"
^
All user defined functions in K are compiled to a bytecode. It can be inspected using !:
!{n:42; x+y+n+1}
("{n:42; x+y+n+1}" / function code itself
0x0268424a67694a214921502100 / bytecode
0xff00020202000b0c090a070800 / bytecode metadata (which local variable to use, reference counting related instructions, etc)
`x`y`n / list of local variables
:: / constants used
42
1)
none.
- Try getting a
'ioerror on your machine. - Now that you know how to debug your K functions, you may want to check out the previous exercises and figure out some of the questions you were not able to finish.