[libc-commits] [libc] [libc]: Implement strfrom* functions and utils (PR #85438)

[Vinayak Dev via libc-commits]

================
@@ -0,0 +1,138 @@
+//===-- Implementation header for strfromx() utilitites -------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// According to the C23 standard, any input character sequences except a
+// precision specifier and the usual floating point formats, namely
+// %{a,A,e,E,f,F,g,G}, are not allowed and any code that does otherwise results
+// in undefined behaviour(including use of a '%%' conversion specifier); which
+// in this case is that the buffer string is simply populated with the format
+// string. The case of the input being NULL should be handled in the calling
+// function (strfromf, strfromd, strfroml) itself.
+
+#ifndef LLVM_LIBC_SRC_STDLIB_STRFROM_UTIL_H
+#define LLVM_LIBC_SRC_STDLIB_STRFROM_UTIL_H
+
+#include "src/__support/CPP/type_traits.h"
+#include "src/__support/str_to_integer.h"
+#include "src/stdio/printf_core/converter_atlas.h"
+#include "src/stdio/printf_core/core_structs.h"
+#include "src/stdio/printf_core/writer.h"
+
+#include <stddef.h>
+
+namespace LIBC_NAMESPACE::internal {
+
+template <typename T>
+using storage_type = typename fputil::FPBits<T>::StorageType;
+
+template <typename T>
+printf_core::FormatSection parse_format_string(const char *__restrict format,
+                                               T fp) {
+  printf_core::FormatSection section;
+  size_t cur_pos = 0;
+
+  // There is no typed conversion function to convert single precision float
+  // to hex exponential format, and the function convert_float_hex_exp()
+  // requires a double or long double value to work correctly.
+  // To work around this, we convert fp to double if it is single precision, and
+  // then use that double precision value in the %{A, a} conversion specifiers.
+  [[maybe_unused]] double new_fp;
+  bool t_is_single_prec_type = cpp::is_same<T, float>::value;
+  if (t_is_single_prec_type)
+    new_fp = (double)fp;
+
+  if (format[cur_pos] == '%') {
+    section.has_conv = true;
+    ++cur_pos;
+
+    // handle precision
+    section.precision = -1;
+    if (format[cur_pos] == '.') {
+      ++cur_pos;
+      section.precision = 0;
+
+      // The standard does not allow the '*' (asterisk) operator for strfromx()
+      // functions
+      if (internal::isdigit(format[cur_pos])) {
+        auto result = internal::strtointeger<int>(format + cur_pos, 10);
+        section.precision += result.value;
+        cur_pos += result.parsed_len;
+      }
+    }
+
+    section.conv_name = format[cur_pos];
+    switch (format[cur_pos]) {
+    case 'a':
+    case 'A':
+      if (t_is_single_prec_type)
+        section.conv_val_raw = cpp::bit_cast<storage_type<double>>(new_fp);
+      else
+        section.conv_val_raw = cpp::bit_cast<storage_type<T>>(fp);
----------------
vinayakdsci wrote:

Yes, but wouldn't I want to do that in strfroml.cpp? I can parse the format string, get the section, and then set the length modifier to L in strfroml.cpp itself. Otherwise we would require a check for all types in the utils file.

https://github.com/llvm/llvm-project/pull/85438