fix "" string interpolation bug

pull/3016/head
Alexander Medvednikov 2019-12-08 20:06:00 +03:00
parent d58982a6f6
commit 4d1f721558
4 changed files with 173 additions and 156 deletions

View File

@ -565,3 +565,11 @@ fn test_inter_before_comp_if() {
}
}
fn test_double_quote_inter() {
a := 1
b := 2
println("${a} ${b}")
assert "${a} ${b}" == "1 2"
assert '${a} ${b}' == "1 2"
}

View File

@ -2280,154 +2280,6 @@ fn format_str(_str string) string {
return str
}
fn (p mut Parser) string_expr() {
is_raw := p.tok == .name && p.lit == 'r'
is_cstr := p.tok == .name && p.lit == 'c'
if is_raw || is_cstr {
p.next()
}
str := p.lit
// No ${}, just return a simple string
if p.peek() != .str_dollar || is_raw {
f := if is_raw { cescaped_path(str) } else { format_str(str) }
// `C.puts('hi')` => `puts("hi");`
/*
Calling a C function sometimes requires a call to a string method
C.fun('ssss'.to_wide()) => fun(string_to_wide(tos3("ssss")))
*/
if (p.calling_c && p.peek() != .dot) || is_cstr || (p.pref.translated && p.mod == 'main') {
p.gen('"$f"')
}
else if p.is_sql {
p.gen("'$str'")
}
else if p.is_js {
p.gen('tos("$f")')
}
else {
p.gen('tos3("$f")')
}
p.next()
return
}
$if js {
p.error('js backend does not support string formatting yet')
}
p.is_alloc = true // $ interpolation means there's allocation
mut args := '"'
mut format := '"'
mut complex_inter := false // for vfmt
for p.tok == .str {
// Add the string between %d's
p.lit = p.lit.replace('%', '%%')
format += format_str(p.lit)
p.next()// skip $
if p.tok != .str_dollar {
continue
}
// Handle .dollar
p.check(.str_dollar)
// If there's no string after current token, it means we are in
// a complex expression (`${...}`)
if p.peek() != .str {
p.fgen('{')
complex_inter = true
}
// Get bool expr inside a temp var
typ, val_ := p.tmp_expr()
val := val_.trim_space()
args += ', $val'
if typ == 'string' {
// args += '.str'
// printf("%.*s", a.len, a.str) syntax
args += '.len, ${val}.str'
}
if typ == 'ustring' {
args += '.len, ${val}.s.str'
}
if typ == 'bool' {
//args += '.len, ${val}.str'
}
// Custom format? ${t.hour:02d}
custom := p.tok == .colon
if custom {
mut cformat := ''
p.next()
if p.tok == .dot {
cformat += '.'
p.next()
}
if p.tok == .minus { // support for left aligned formatting
cformat += '-'
p.next()
}
cformat += p.lit// 02
p.next()
fspec := p.lit // f
cformat += fspec
if fspec == 's' {
//println('custom str F=$cformat | format_specifier: "$fspec" | typ: $typ ')
if typ != 'string' {
p.error('only V strings can be formatted with a :${cformat} format, but you have given "${val}", which has type ${typ}')
}
args = args.all_before_last('${val}.len, ${val}.str') + '${val}.str'
}
format += '%$cformat'
p.next()
}
else {
f := p.typ_to_fmt(typ, 0)
if f == '' {
is_array := typ.starts_with('array_')
typ2 := p.table.find_type(typ)
has_str_method := p.table.type_has_method(typ2, 'str')
if is_array || has_str_method {
if is_array && !has_str_method {
p.gen_array_str(typ2)
}
tmp_var := p.get_tmp()
p.cgen.insert_before('string $tmp_var = ${typ}_str(${val});')
args = args.all_before_last(val) + '${tmp_var}.len, ${tmp_var}.str'
format += '%.*s '
}
else {
p.error('unhandled sprintf format "$typ" ')
}
}
format += f
}
//println('interpolation format is: |${format}| args are: |${args}| ')
}
if complex_inter {
p.fgen('}')
}
//p.fgen('\'')
// println("hello %d", num) optimization.
if p.cgen.nogen {
return
}
// println: don't allocate a new string, just print it.
$if !windows {
cur_line := p.cgen.cur_line.trim_space()
if cur_line == 'println (' && p.tok != .plus {
p.cgen.resetln(cur_line.replace('println (', 'printf('))
p.gen('$format\\n$args')
return
}
}
// '$age'! means the user wants this to be a tmp string (uses global buffer, no allocation,
// won't be used again)
// TODO remove this hack, do this automatically
if p.tok == .not {
p.check(.not)
p.gen('_STR_TMP($format$args)')
}
else {
// Otherwise do len counting + allocation + sprintf
p.gen('_STR($format$args)')
}
}
// m := map[string]int{}
// m := { 'one': 1 }
fn (p mut Parser) map_init() string {

View File

@ -298,14 +298,15 @@ fn (s mut Scanner) scan() ScanRes {
s.inside_string = false
}
}
// end of `$expr`
if s.inter_start && next_char != `.` {
s.inter_end = true
s.inter_start = false
}
if s.pos == 0 && next_char == ` ` {
s.pos++
//If a single letter name at the start of the file, increment
//Otherwise the scanner would be stuck at s.pos = 0
s.pos++
}
return scan_res(.name, name)
}
@ -390,8 +391,7 @@ fn (s mut Scanner) scan() ScanRes {
return scan_res(.lcbr, '')
}
`$` {
// if s.inter_start {
if s.inside_string {// || s.inter_start {
if s.inside_string {
return scan_res(.str_dollar, '')
} else {
return scan_res(.dollar, '')
@ -402,8 +402,7 @@ fn (s mut Scanner) scan() ScanRes {
// s = `hello ${name} !`
if s.inside_string {
s.pos++
// TODO UNNEEDED?
if s.text[s.pos] == single_quote {
if s.text[s.pos] == s.quote {
s.inside_string = false
return scan_res(.str, '')
}
@ -704,15 +703,19 @@ fn (s mut Scanner) ident_string() string {
if c == `0` && s.pos > 5 && s.expect('\\x0', s.pos - 3) {
s.error('0 character in a string literal')
}
// ${var}
if c == `{` && prevc == `$` && !is_raw && !s.is_fmt && s.count_symbol_before(s.pos-2, slash) % 2 == 0 {
// ${var} (ignore in vfmt mode)
if c == `{` && prevc == `$` && !is_raw && !s.is_fmt &&
s.count_symbol_before(s.pos-2, slash) % 2 == 0
{
s.inside_string = true
// so that s.pos points to $ at the next step
s.pos -= 2
break
}
// $var
if (c.is_letter() || c == `_`) && prevc == `$` && !s.is_fmt && !is_raw && s.count_symbol_before(s.pos-2, slash) % 2 == 0 {
if (c.is_letter() || c == `_`) && prevc == `$` && !s.is_fmt &&
!is_raw && s.count_symbol_before(s.pos-2, slash) % 2 == 0
{
s.inside_string = true
s.inter_start = true
s.pos -= 2

View File

@ -0,0 +1,154 @@
// Copyright (c) 2019 Alexander Medvednikov. All rights reserved.
// Use of this source code is governed by an MIT license
// that can be found in the LICENSE file.
module compiler
fn (p mut Parser) string_expr() {
is_raw := p.tok == .name && p.lit == 'r'
is_cstr := p.tok == .name && p.lit == 'c'
if is_raw || is_cstr {
p.next()
}
str := p.lit
// No ${}, just return a simple string
if p.peek() != .str_dollar || is_raw {
f := if is_raw { cescaped_path(str) } else { format_str(str) }
// `C.puts('hi')` => `puts("hi");`
/*
Calling a C function sometimes requires a call to a string method
C.fun('ssss'.to_wide()) => fun(string_to_wide(tos3("ssss")))
*/
if (p.calling_c && p.peek() != .dot) || is_cstr || (p.pref.translated && p.mod == 'main') {
p.gen('"$f"')
}
else if p.is_sql {
p.gen("'$str'")
}
else if p.is_js {
p.gen('tos("$f")')
}
else {
p.gen('tos3("$f")')
}
p.next()
return
}
$if js {
p.error('js backend does not support string formatting yet')
}
p.is_alloc = true // $ interpolation means there's allocation
mut args := '"'
mut format := '"'
mut complex_inter := false // for vfmt
for p.tok == .str {
// Add the string between %d's
p.lit = p.lit.replace('%', '%%')
format += format_str(p.lit)
p.next()// skip $
if p.tok != .str_dollar {
continue
}
// Handle .dollar
p.check(.str_dollar)
// If there's no string after current token, it means we are in
// a complex expression (`${...}`)
if p.peek() != .str {
p.fgen('{')
complex_inter = true
}
// Get bool expr inside a temp var
typ, val_ := p.tmp_expr()
val := val_.trim_space()
args += ', $val'
if typ == 'string' {
// args += '.str'
// printf("%.*s", a.len, a.str) syntax
args += '.len, ${val}.str'
}
if typ == 'ustring' {
args += '.len, ${val}.s.str'
}
if typ == 'bool' {
//args += '.len, ${val}.str'
}
// Custom format? ${t.hour:02d}
custom := p.tok == .colon
if custom {
mut cformat := ''
p.next()
if p.tok == .dot {
cformat += '.'
p.next()
}
if p.tok == .minus { // support for left aligned formatting
cformat += '-'
p.next()
}
cformat += p.lit// 02
p.next()
fspec := p.lit // f
cformat += fspec
if fspec == 's' {
//println('custom str F=$cformat | format_specifier: "$fspec" | typ: $typ ')
if typ != 'string' {
p.error('only V strings can be formatted with a :${cformat} format, but you have given "${val}", which has type ${typ}')
}
args = args.all_before_last('${val}.len, ${val}.str') + '${val}.str'
}
format += '%$cformat'
p.next()
}
else {
f := p.typ_to_fmt(typ, 0)
if f == '' {
is_array := typ.starts_with('array_')
typ2 := p.table.find_type(typ)
has_str_method := p.table.type_has_method(typ2, 'str')
if is_array || has_str_method {
if is_array && !has_str_method {
p.gen_array_str(typ2)
}
tmp_var := p.get_tmp()
p.cgen.insert_before('string $tmp_var = ${typ}_str(${val});')
args = args.all_before_last(val) + '${tmp_var}.len, ${tmp_var}.str'
format += '%.*s '
}
else {
p.error('unhandled sprintf format "$typ" ')
}
}
format += f
}
//println('interpolation format is: |${format}| args are: |${args}| ')
}
if complex_inter {
p.fgen('}')
}
//p.fgen('\'')
// println("hello %d", num) optimization.
if p.cgen.nogen {
return
}
// println: don't allocate a new string, just print it.
$if !windows {
cur_line := p.cgen.cur_line.trim_space()
if cur_line == 'println (' && p.tok != .plus {
p.cgen.resetln(cur_line.replace('println (', 'printf('))
p.gen('$format\\n$args')
return
}
}
// '$age'! means the user wants this to be a tmp string (uses global buffer, no allocation,
// won't be used again)
// TODO remove this hack, do this automatically
if p.tok == .not {
p.check(.not)
p.gen('_STR_TMP($format$args)')
}
else {
// Otherwise do len counting + allocation + sprintf
p.gen('_STR($format$args)')
}
}