improved private tag parsing

This commit is contained in:
Folkert 2021-03-10 17:26:42 +01:00
parent edd54ab4ab
commit ea32a37315
4 changed files with 242 additions and 70 deletions

View file

@ -299,25 +299,33 @@ pub enum BadIdent {
WeirdDotQualified(Row, Col),
DoubleDot(Row, Col),
StrayDot(Row, Col),
StrayAt(Row, Col),
BadPrivateTag(Row, Col),
}
/// Parse an identifier into a string.
///
/// This is separate from the `ident` Parser because string interpolation
/// wants to use it this way.
fn chomp_lowercase_part(buffer: &[u8]) -> Result<&str, Progress> {
chomp_part(|c: char| c.is_lowercase(), buffer)
}
/// a `.foo` accessor function
fn chomp_accessor(buffer: &[u8], row: Row, col: Col) -> Result<&str, BadIdent> {
fn chomp_uppercase_part(buffer: &[u8]) -> Result<&str, Progress> {
chomp_part(|c: char| c.is_uppercase(), buffer)
}
#[inline(always)]
fn chomp_part<F>(leading_is_good: F, buffer: &[u8]) -> Result<&str, Progress>
where
F: Fn(char) -> bool,
{
// assumes the leading `.` has been chomped already
use encode_unicode::CharExt;
let mut chomped = 0;
if let Ok((ch, width)) = char::from_utf8_slice_start(&buffer[chomped..]) {
if ch.is_lowercase() {
if leading_is_good(ch) {
chomped += width;
} else {
return Err(BadIdent::StrayDot(row, col + 1));
return Err(NoProgress);
}
}
@ -331,17 +339,56 @@ fn chomp_accessor(buffer: &[u8], row: Row, col: Col) -> Result<&str, BadIdent> {
}
if chomped == 0 {
Err(BadIdent::StrayDot(row, col + 1))
} else if let Ok(('.', _)) = char::from_utf8_slice_start(&buffer[chomped..]) {
Err(BadIdent::WeirdAccessor(row, col))
Err(NoProgress)
} else {
let name = unsafe { std::str::from_utf8_unchecked(&buffer[..chomped]) };
dbg!(name);
Ok(name)
}
}
/// a `.foo` accessor function
fn chomp_accessor(buffer: &[u8], row: Row, col: Col) -> Result<&str, BadIdent> {
// assumes the leading `.` has been chomped already
use encode_unicode::CharExt;
match chomp_lowercase_part(buffer) {
Ok(name) => {
let chomped = name.len();
if let Ok(('.', _)) = char::from_utf8_slice_start(&buffer[chomped..]) {
Err(BadIdent::WeirdAccessor(row, col))
} else {
Ok(name)
}
}
Err(_) => {
// we've already made progress with the initial `.`
Err(BadIdent::StrayDot(row, col + 1))
}
}
}
/// a `@Token` private tag
fn chomp_private_tag(buffer: &[u8], row: Row, col: Col) -> Result<&str, BadIdent> {
// assumes the leading `@` has NOT been chomped already
debug_assert_eq!(buffer.get(0), Some(&b'@'));
use encode_unicode::CharExt;
match chomp_uppercase_part(&buffer[1..]) {
Ok(name) => {
let chomped = 1 + name.len();
if let Ok(('.', _)) = char::from_utf8_slice_start(&buffer[chomped..]) {
Err(BadIdent::BadPrivateTag(row, col + chomped as u16))
} else {
Ok(name)
}
}
Err(_) => Err(BadIdent::BadPrivateTag(row, col + 1)),
}
}
fn parse_ident_help_help<'a>(
arena: &'a Bump,
mut state: State<'a>,
@ -350,7 +397,6 @@ fn parse_ident_help_help<'a>(
let mut noncapitalized_parts: Vec<&'a str> = Vec::new_in(arena);
let mut is_capitalized;
let is_accessor_fn;
let mut is_private_tag = false;
let bytes = state.bytes;
let mut chomped_capitalized = 0;
@ -387,35 +433,17 @@ fn parse_ident_help_help<'a>(
Err(fail) => return Err((MadeProgress, fail, state)),
}
} else if first_ch == '@' {
state = advance_state!(state, bytes_parsed)?;
match chomp_private_tag(state.bytes, state.line, state.column) {
Ok(tagname) => {
let bytes_parsed = 1 + tagname.len();
// '@' must always be followed by a capital letter!
match peek_utf8_char(&state) {
Ok((next_ch, next_bytes_parsed)) => {
if next_ch.is_uppercase() {
state = advance_state!(state, next_bytes_parsed)?;
state = advance_state!(state, bytes_parsed)?;
part_buf.push('@');
part_buf.push(next_ch);
chomped_part_buf += 1 + next_bytes_parsed;
is_private_tag = true;
is_capitalized = true;
is_accessor_fn = false;
} else {
return Err((
MadeProgress,
BadIdent::PrivateTagNotUppercase(state.line, state.column),
state,
));
}
return Ok((MadeProgress, (Ident::PrivateTag(tagname), None), state));
}
Err(_reason) => {
return Err((
MadeProgress,
BadIdent::PrivateTagNotUppercase(state.line, state.column),
state,
));
Err(fail) => {
state = advance_state!(state, 1)?;
return Err((MadeProgress, fail, state));
}
}
} else {
@ -555,7 +583,7 @@ fn parse_ident_help_help<'a>(
let answer = if is_accessor_fn {
// Handle accessor functions first because they have the strictest requirements.
// Accessor functions may have exactly 1 noncapitalized part, and no capitalzed parts.
if cparts == 0 && noncapitalized_parts.len() == 1 && !is_private_tag {
if cparts == 0 && noncapitalized_parts.len() == 1 {
// an accessor starts with a `.`, but we drop that from the name
let value = unsafe {
std::str::from_utf8_unchecked(&bytes[1 + chomped..1 + chomped + chomped_part_buf])
@ -585,11 +613,7 @@ fn parse_ident_help_help<'a>(
1 => {
let chomped = chomped_capitalized;
let value = unsafe { std::str::from_utf8_unchecked(&bytes[..chomped]) };
if is_private_tag {
Ident::PrivateTag(value)
} else {
Ident::GlobalTag(value)
}
Ident::GlobalTag(value)
}
_ => {
// This is a qualified tag, which is not allowed!
@ -600,13 +624,6 @@ fn parse_ident_help_help<'a>(
));
}
}
} else if is_private_tag {
// This is qualified field access with an '@' in front, which does not make sense!
return Err((
MadeProgress,
BadIdent::PrivateTagFieldAccess(state.line, state.column),
state,
));
} else {
// We have multiple noncapitalized parts, so this must be field access.
let module_name = if cparts == 0 {

View file

@ -1,4 +1,5 @@
use roc_collections::all::MutSet;
use roc_parse::parser::{Col, Row};
use roc_problem::can::PrecedenceProblem::BothNonAssociative;
use roc_problem::can::{FloatErrorKind, IntErrorKind, Problem, RuntimeError};
use roc_region::all::Region;
@ -450,14 +451,106 @@ fn to_bad_ident_expr_report<'b>(
])
}
PrivateTagFieldAccess(_row, _col) => alloc.stack(vec![
alloc.reflow("I am very confused by this field access:"),
alloc.region(surroundings),
alloc.concat(vec![
alloc.reflow(r"It looks like a record field access on a private tag.")
]),
]),
_ => todo!(),
PrivateTagFieldAccess(row, col) => {
let region =
Region::from_rows_cols(surroundings.start_line, surroundings.start_col, row, col);
alloc.stack(vec![
alloc.reflow("I am very confused by this field access:"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![
alloc.reflow(r"It looks like a record field access on a private tag.")
]),
])
}
Underscore(row, col) => {
let region =
Region::from_rows_cols(surroundings.start_line, surroundings.start_col, row, col);
alloc.stack(vec![
alloc.reflow("Underscores are not allowed in identifier names:"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![alloc.reflow(
r"I recommend using camelCase, it is the standard in the Roc ecosystem.",
)]),
])
}
DoubleDot(row, col) => {
let region =
Region::from_rows_cols(surroundings.start_line, surroundings.start_col, row, col);
alloc.stack(vec![
alloc.reflow("I am very confused by these two dots in a row:"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![
alloc.reflow(r"There always needs to be a name after a dot.")
]),
])
}
StrayAt(row, col) => {
let region =
Region::from_rows_cols(surroundings.start_line, surroundings.start_col, row, col);
alloc.stack(vec![
alloc.reflow("I am very confused by this @ symbol"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![alloc.reflow(r"I expected a private tag.")]),
])
}
BadPrivateTag(row, col) => {
use BadIdentNext::*;
match what_is_next(alloc.src_lines, row, col) {
LowercaseAccess(width) => {
let region = Region::from_rows_cols(row, col, row, col + width);
alloc.stack(vec![
alloc.reflow("I am very confused by this field access:"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![
alloc.reflow(r"It looks like a record field access on a private tag.")
]),
])
}
UppercaseAccess(width) => {
let region = Region::from_rows_cols(row, col, row, col + width);
alloc.stack(vec![
alloc.reflow("I am very confused by this expression:"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![
alloc.reflow(
r"Looks like a private tag is treated like a module name. ",
),
alloc.reflow(r"Maybe you wanted a qualified name, like "),
alloc.parser_suggestion("Json.Decode.string"),
alloc.text("?"),
]),
])
}
Other(Some(c)) if c.is_lowercase() => {
let region = Region::from_rows_cols(
surroundings.start_line,
surroundings.start_col + 1,
row,
col + 1,
);
alloc.stack(vec![
alloc.reflow("I am trying to parse a private tag here:"),
alloc.region_with_subregion(surroundings, region),
alloc.concat(vec![
alloc.reflow(r"But after the "),
alloc.keyword("@"),
alloc.reflow(r" symbol I found a lowercase letter. "),
alloc.reflow(r"All tag names (global and private)"),
alloc.reflow(r" must start with an uppercase letter, like "),
alloc.parser_suggestion("@UUID"),
alloc.reflow(" or "),
alloc.parser_suggestion("@Secrets"),
alloc.reflow("."),
]),
])
}
other => todo!("{:?}", other),
}
}
}
}
@ -591,6 +684,69 @@ fn to_bad_ident_pattern_report<'b>(
}
}
#[derive(Debug)]
enum BadIdentNext<'a> {
LowercaseAccess(u16),
UppercaseAccess(u16),
NumberAccess(u16),
Keyword(&'a str),
DanglingDot,
Other(Option<char>),
}
fn what_is_next<'a>(source_lines: &'a [&'a str], row: Row, col: Col) -> BadIdentNext<'a> {
let row_index = row as usize;
let col_index = col as usize;
match source_lines.get(row_index) {
None => BadIdentNext::Other(None),
Some(line) => {
let chars = &line[col_index..];
let mut it = chars.chars();
match roc_parse::keyword::KEYWORDS
.iter()
.find(|keyword| crate::error::parse::starts_with_keyword(chars, keyword))
{
Some(keyword) => BadIdentNext::Keyword(keyword),
None => match it.next() {
None => BadIdentNext::Other(None),
Some('.') => match it.next() {
Some(c) if c.is_lowercase() => {
BadIdentNext::LowercaseAccess(2 + till_whitespace(it) as u16)
}
Some(c) if c.is_uppercase() => {
BadIdentNext::UppercaseAccess(2 + till_whitespace(it) as u16)
}
Some(c) if c.is_ascii_digit() => {
BadIdentNext::NumberAccess(2 + till_whitespace(it) as u16)
}
_ => BadIdentNext::DanglingDot,
},
Some(c) => BadIdentNext::Other(Some(c)),
},
}
}
}
}
fn till_whitespace<I>(mut it: I) -> usize
where
I: Iterator<Item = char>,
{
let mut chomped = 0;
while let Some(c) = it.next() {
if c.is_ascii_whitespace() || c == '#' {
break;
} else {
chomped += 1;
continue;
}
}
chomped
}
fn pretty_runtime_error<'b>(
alloc: &'b RocDocAllocator<'b>,
runtime_error: RuntimeError,

View file

@ -3012,7 +3012,7 @@ fn what_is_next<'a>(source_lines: &'a [&'a str], row: Row, col: Col) -> Next<'a>
}
}
fn starts_with_keyword(rest_of_line: &str, keyword: &str) -> bool {
pub fn starts_with_keyword(rest_of_line: &str, keyword: &str) -> bool {
if let Some(stripped) = rest_of_line.strip_prefix(keyword) {
match stripped.chars().next() {
None => true,

View file

@ -4148,14 +4148,13 @@ mod test_reporting {
r#"
SYNTAX PROBLEM
I am trying to parse a qualified name here:
I am very confused by this expression:
1 @Foo.Bar
^
^^^^
This looks like a qualified tag name to me, but tags cannot be
qualified! Maybe you wanted a qualified name, something like
Json.Decode.string?
Looks like a private tag is treated like a module name. Maybe you
wanted a qualified name, like Json.Decode.string?
"#
),
)
@ -5523,7 +5522,7 @@ mod test_reporting {
I am very confused by this field access:
1 @UUID.bar
^^^^^^^^^
^^^^
It looks like a record field access on a private tag.
"#