parser: Add source position tracking of nodes

[mewmew]

From ast.go:

// TODO: Add source position tracking of nodes.

This requires some careful considerations. How would we like to store the node source position information? Should we always store both the start and the end of a node, or just the start and calculate the end from the length of the node (would this always be possible?).

Related to issue #22 which tracks the pretty-printing of line:col pairs.

[mewmew]

Gocc already provides the positional information when generating error messages for input streams which are not reducible by the grammar. We may therefore postpone the implementation of this feature until the semantic analysis stage, at which point positional information should be present for each node of the AST.

[mewmew]

Positional information was added to each AST node with commit 313ddfd0213a62830974dad23bd2305b5c4d3758.

Tracking of positional information was added in commit 5080451b75dffccc711931fce047bcffd5212938. Note, this commit breaks all parser test cases since they don't contain any positional information yet.

Commit 2ba2f0f5b2a65c38170365dffb221cac65fd5ff3 added positional information to the quiet/lexer/l05.c test case. All other parser test cases are still broken.

[mewmew]

Updated by commit 17205c0120c2046608b351f73f0a9cdecb52ffec.

[sangisos]

Updated by commit b64e65eea7f70924661f2e91e1ec0affa47f21df

[mewmew]

As far as I can tell, this issue was resolved with commit b64e65e. Anything else we need before closing it?

The conversion of positional information to line:col pairs is tracked by issue #25.

[sangisos]

Well it depends on if we want to clean up unnecessary to track positions, and just track the essential ones.

[mewmew]

Well it depends on if we want to clean up unnecessary to track positions, and just track the essential ones.

That's true. Lets give it a couple of days to sink in before we take a decision. Basically, once we finish the semantic analysis phase, we will have a better understanding of which parts of the AST API we are actually using. At that point, perhaps we'll get rid of the rest to optimize memory usage.

[mewmew]

Lets discuss this issue next time we meet. Which positional tracking information should we include in the nodes? The minimum required for good semantic messages, the current information, or all information of token positions?

[mewmew]

To facilitate generation of informative error messages, a decision has been taken to store positional information for each relevant token.

An example from clang is presented below, where the right-parenthesis ) is pointed out in the error message.

Contents of a.c:

void f(int a, int b, int c) {
}

int main(void) {
    f(1, 2, 3, 4, 5);
}

u@x1 ~/Desktop> clang -o a a.c
a.c:5:13: error: too many arguments to function call, expected 3, have 5
        f(1, 2, 3, 4, 5);
        ~          ^~~~
a.c:1:1: note: 'f' declared here
void f(int a, int b, int c) {
^
1 error generated.

Closing this issue for now.