Add next_word() overload that doesn't skip consecutive delimiters

In particular, this version can be used to parse lines while observing
blanks.

4 files changed, 135 insertions, 0 deletions

diff --git a/libbutl/utility.hxx b/libbutl/utility.hxx
index a129276..779a0aa 100644
--- a/libbutl/utility.hxx
+++ b/libbutl/utility.hxx
@@ -190,6 +190,24 @@ namespace butl
   //
   // The second version examines up to the n'th character in the string.
   //
+  // The third version, instead of skipping consecutive delimiters, treats
+  // them as separating empty words. The additional m variable contains an
+  // unspecified internal state and should be initialized to 0. Note that in
+  // this case you should use the (b == n) condition to detect the end. Note
+  // also that a leading delimiter is considered as separating an empty word
+  // from the rest and the trailing delimiter is considered as separating the
+  // rest from an empty word. For example, this is how to parse lines while
+  // observing blanks:
+  //
+  // for (size_t b (0), e (0), m (0), n (s.size ());
+  //      next_word (s, n, b, e, m, '\n', '\r'), b != n; )
+  // {
+  //   string l (s, b, e - b);
+  // }
+  //
+  // For string "\na\n" this code will observe the {"", "a", ""} words. And
+  // for just "\n" it will observe the {"", ""} words.
+  //
   std::size_t
   next_word (const std::string&, std::size_t& b, std::size_t& e,
              char d1 = ' ', char d2 = '\0');
@@ -198,6 +216,11 @@ namespace butl
   next_word (const std::string&, std::size_t n, std::size_t& b, std::size_t& e,
              char d1 = ' ', char d2 = '\0');
 
+  std::size_t
+  next_word (const std::string&, std::size_t n,
+             std::size_t& b, std::size_t& e, std::size_t& m,
+             char d1 = ' ', char d2 = '\0');
+
   // Sanitize a string to only contain characters valid in an identifier
   // (ASCII alphanumeric plus `_`) replacing all others with `_`.
   //
diff --git a/libbutl/utility.ixx b/libbutl/utility.ixx
index 0ce33a7..fda1ce5 100644
--- a/libbutl/utility.ixx
+++ b/libbutl/utility.ixx
@@ -210,6 +210,66 @@ namespace butl
     return e - b;
   }
 
+  inline std::size_t
+  next_word (const std::string& s,
+             std::size_t n, std::size_t& b, std::size_t& e, std::size_t& m,
+             char d1, char d2)
+  {
+    // An empty word will necessarily be represented as b and e being the
+    // position of a delimiter. Consider these corner cases (in all three we
+    // should produce two words):
+    //
+    // \n
+    // a\n
+    // \na
+    //
+    // It feels sensible to represent an empty word as the position of the
+    // trailing delimiter except if it is the last character (the first two
+    // cases). Thus the additional m state, which, if 0 or 1 indicates the
+    // number of delimiters to skip before parsing the next word and 2 if
+    // this is a trailing delimiter for which we need to fake an empty word
+    // with the leading delimiter.
+
+    if (b != e)
+      b = e;
+
+    if (m > 1)
+    {
+      --m;
+      return 0;
+    }
+
+    // Skip the leading delimiter, if any.
+    //
+    b += m;
+
+    if (b == n)
+    {
+      e = n;
+      return 0;
+    }
+
+    // Find first trailing delimiter.
+    //
+    m = 0;
+    for (e = b; e != n; ++e)
+    {
+      if (s[e] == d1 || s[e] == d2)
+      {
+        m = 1;
+
+        // Handle the special delimiter as the last character case.
+        //
+        if (e + 1 == n)
+          ++m;
+
+        break;
+      }
+    }
+
+    return e - b;
+  }
+
   inline std::string&
   sanitize_identifier (std::string& s)
   {
diff --git a/tests/next-word/buildfile b/tests/next-word/buildfile
new file mode 100644
index 0000000..e06cd88
--- /dev/null
+++ b/tests/next-word/buildfile
@@ -0,0 +1,6 @@
+# file      : tests/next-word/buildfile
+# license   : MIT; see accompanying LICENSE file
+
+import libs = libbutl%lib{butl}
+
+exe{driver}: {hxx cxx}{*} $libs
diff --git a/tests/next-word/driver.cxx b/tests/next-word/driver.cxx
new file mode 100644
index 0000000..4ebe1a5
--- /dev/null
+++ b/tests/next-word/driver.cxx
@@ -0,0 +1,46 @@
+// file      : tests/next-word/driver.cxx -*- C++ -*-
+// license   : MIT; see accompanying LICENSE file
+
+#include <vector>
+#include <string>
+//#include <iostream>
+
+#include <libbutl/utility.hxx>
+
+#undef NDEBUG
+#include <cassert>
+
+using namespace std;
+using namespace butl;
+
+using strings = vector<string>;
+
+static strings
+parse_lines (const string& s)
+{
+  strings r;
+  for (size_t b (0), e (0), m (0), n (s.size ());
+       next_word (s, n, b, e, m, '\n', '\r'), b != n; )
+  {
+    //cerr << "'" << string (s, b, e - b) << "'" << endl;
+    r.push_back (string (s, b, e - b));
+  }
+  return r;
+}
+
+int
+main ()
+{
+  assert ((parse_lines("") == strings {}));
+  assert ((parse_lines("a") == strings {"a"}));
+  assert ((parse_lines("\n") == strings {"", ""}));
+  assert ((parse_lines("\n\n") == strings {"", "", ""}));
+  assert ((parse_lines("\n\n\n") == strings {"", "", "", ""}));
+  assert ((parse_lines("\na") == strings {"", "a"}));
+  assert ((parse_lines("\n\na") == strings {"", "", "a"}));
+  assert ((parse_lines("a\n") == strings {"a", ""}));
+  assert ((parse_lines("a\n\n") == strings {"a", "", ""}));
+  assert ((parse_lines("a\nb") == strings {"a", "b"}));
+  assert ((parse_lines("a\n\nb") == strings {"a", "", "b"}));
+  assert ((parse_lines("\na\nb\n") == strings {"", "a", "b", ""}));
+}