Building Funa Pt.2: Parser Edge Cases

Last week I continued work on the parser and decided to focus on finding a good flow at the start. My plan is that when I establish a good flow with one part, the rest would be easier. I focused on the <var-bind> statement:

<program> ::= <statement>*

<statement> ::= <var-bind>
              | <fn-def>

<var-bind> ::= LET IDENT ASSIGNMENT <expr>

<expr> ::= NUMBER | STRING

<fn-def> ::= FUNCTION IDENT LPAREN <param-list>? RPAREN COLON <type> <block>

<param-list> ::= <param> ( COMMA <param> )*
<param> ::= IDENT COLON <type>

<block> ::= LBRACE <statement>* <return-stmt> RBRACE

<return-stmt> ::= RETURN <expr>

<type> ::= "integer" | "string" | "float"

What is a good flow? Well, the parser will translate the language grammar into a syntax tree. The grammar definition for the syntax of statements may be different, but the processes are similar. Here are two important consistencies when building the tree:

1. Syntax error handling.
2. Token peeking for construction the Abstract Syntax Tree. Like a mini state machine.

They were both simple to write, especially the second one. For the syntax error handling, I had some edge cases. It is the classic off by one error. Let me give you some context.

Error on the last line on the last column

Funa code:

let some = 4
let some = 2
let some = 1
let some =

Output:

Syntax Error on line 4, column 10: expected ["INTEGER", "FLOAT", "STRING"], got EOF

Comment:

Line 4 is correct. Column 10 is correct.

Error in the middle

Funa code:

let some = 4
let = 2
let some = 1
let some = 0

Output:

Syntax Error on line 2, column 5: expected ["IDENTIFIER"], got ASSIGNMENT

Comment:

Line 2 is correct. Column 5 is correct.

Our so called edge case: error on first line, first column

Funa code:

et some = 4
let some = 2
let some = 1
let some = 0

Output:

Syntax Error on line 1, column 4: expected ["LET", "FUNCTION"], got IDENTIFIER

Comment:

Line 1 is correct. Column 4, not so much. I was expecting it to point to column one. It is consistent if the syntax error ends up being on the first column no matter the line.

---

I had two thoughts:

1. Continue to spend time figuring out why that edge case exists.
2. Remove the column number tracker entirely.

I decided to leave it as it is and focus on the main things. It may not be a big deal as I make it out to be. If I eventually find out how to deal with it, or if anyone figures it out, feel free to make a pull request on the repo.

Concluding Words

That was the progress made with regards to learning Odin and building funa. I won’t say I learnt anything syntax-wise to Odin. Its syntax is easier to grasp. I do enjoy the manual memmory management. It’s not as scary as we make it.