Add tab_parser

3 files changed, 221 insertions, 0 deletions

diff --git a/butl/buildfile b/butl/buildfile
index 0aa0aae..b54a887 100644
--- a/butl/buildfile
+++ b/butl/buildfile
@@ -25,6 +25,7 @@ lib{butl}:                                  \
    {hxx         cxx}{ sha256              } \
    {hxx            }{ small-vector        } \
    {hxx     txx    }{ string-table        } \
+   {hxx         cxx}{ tab-parser          } \
    {hxx         cxx}{ target-triplet      } \
    {hxx         cxx}{ timestamp           } \
    {hxx ixx     cxx}{ utility             } \
diff --git a/butl/tab-parser b/butl/tab-parser
new file mode 100644
index 0000000..6aa0705
--- /dev/null
+++ b/butl/tab-parser
@@ -0,0 +1,78 @@
+// file      : butl/tab-parser -*- C++ -*-
+// copyright : Copyright (c) 2014-2017 Code Synthesis Ltd
+// license   : MIT; see accompanying LICENSE file
+
+#ifndef BUTL_TAB_PARSER
+#define BUTL_TAB_PARSER
+
+#include <string>
+#include <vector>
+#include <iosfwd>
+#include <cstdint>   // uint64_t
+#include <stdexcept> // runtime_error
+
+#include <butl/export>
+
+#include <butl/char-scanner>
+
+namespace butl
+{
+  class LIBBUTL_EXPORT tab_parsing: public std::runtime_error
+  {
+  public:
+    tab_parsing (const std::string& name,
+                 std::uint64_t line,
+                 std::uint64_t column,
+                 const std::string& description);
+
+    std::string name;
+    std::uint64_t line;
+    std::uint64_t column;
+    std::string description;
+  };
+
+  // Line and columns are useful for issuing diagnostics about invalid or
+  // missing fields.
+  //
+  struct tab_field
+  {
+    std::string value;    // Field string (quoting preserved).
+    std::uint64_t column; // Field start column number (one-based).
+  };
+
+  struct tab_fields: std::vector<tab_field>
+  {
+    std::uint64_t line;       // Line number (one-based).
+    std::uint64_t end_column; // End-of-line column (line length).
+  };
+
+  // Read and parse lines consisting of space-separated fields. Field can
+  // contain single or double quoted substrings (with spaces) which are
+  // interpreted but preserved. No escaping of the quote characters is
+  // supported. Blank lines and lines that start with # (collectively called
+  // empty lines) are ignored.
+  //
+  class LIBBUTL_EXPORT tab_parser: protected butl::char_scanner
+  {
+  public:
+    tab_parser (std::istream& is, const std::string& name)
+        : char_scanner (is), name_ (name) {}
+
+    // Return next line of fields. Skip empty lines. Empty result denotes the
+    // end of stream.
+    //
+    tab_fields
+    next ();
+
+  private:
+    // Skip spaces and return the first peeked non-space character.
+    //
+    xchar
+    skip_spaces ();
+
+  private:
+    const std::string name_;
+  };
+}
+
+#endif // BUTL_TAB_PARSER
diff --git a/butl/tab-parser.cxx b/butl/tab-parser.cxx
new file mode 100644
index 0000000..bae9327
--- /dev/null
+++ b/butl/tab-parser.cxx
@@ -0,0 +1,142 @@
+// file      : butl/tab-parser.cxx -*- C++ -*-
+// copyright : Copyright (c) 2014-2017 Code Synthesis Ltd
+// license   : MIT; see accompanying LICENSE file
+
+#include <butl/tab-parser>
+
+#include <cassert>
+#include <sstream>
+
+using namespace std;
+
+namespace butl
+{
+  using parsing = tab_parsing;
+
+  // tab_parser
+  //
+  tab_fields tab_parser::
+  next ()
+  {
+    tab_fields r;
+    xchar c (skip_spaces ()); // Skip empty lines and leading spaces.
+
+    auto eol   = [&c] () -> bool {return eos (c) || c == '\n';};
+    auto space = [&c] () -> bool {return c == ' ' || c == '\t';};
+    auto next  = [&c, this] () {get (); c = peek ();};
+
+    r.line = c.line;
+
+    // Read line fields until eos or the newline character.
+    //
+    while (!eol ())
+    {
+      for (; !eol () && space (); next ()) ; // Skip space characters.
+
+      if (eol ()) // No more fields.
+        break;
+
+      // Read the field. Here we scan until the first whitespace character that
+      // appears out of quotes.
+      //
+      tab_field tf ({string (), c.column});
+      char quoting ('\0'); // Current quoting mode, can be used as bool.
+
+      for (; !eol (); next ())
+      {
+        if (!quoting)
+        {
+          if (space ())                   // End of the field.
+            break;
+          else if (c == '"' || c == '\'') // Begin of quoted string.
+            quoting = c;
+        }
+        else if (c == quoting)            // End of quoted string.
+          quoting = '\0';
+
+        tf.value += c;
+      }
+
+      if (quoting)
+        throw parsing (name_, c.line, c.column, "unterminated quoted string");
+
+      r.emplace_back (move (tf));
+    }
+
+    r.end_column = c.column;
+
+    // Read out eof or newline character from the stream. Note that "reading"
+    // eof multiple times is safe.
+    //
+    get ();
+    return r;
+  }
+
+  tab_parser::xchar tab_parser::
+  skip_spaces ()
+  {
+    xchar c (peek ());
+    bool start (c.column == 1);
+
+    for (; !eos (c); c = peek ())
+    {
+      switch (c)
+      {
+      case ' ':
+      case '\t':
+        break;
+      case '\n':
+        {
+          // Skip empty lines.
+          //
+          if (!start)
+            return c;
+
+          break;
+        }
+      case '#':
+        {
+          // We only recognize '#' as a start of a comment at the beginning
+          // of the line (sans leading spaces).
+          //
+          if (!start)
+            return c;
+
+          get ();
+
+          // Read until newline or eos.
+          //
+          for (c = peek (); !eos (c) && c != '\n'; c = peek ())
+            get ();
+
+          continue;
+        }
+      default:
+        return c; // Not a space.
+      }
+
+      get ();
+    }
+
+    return c;
+  }
+
+  // tab_parsing
+  //
+  static string
+  format (const string& n, uint64_t l, uint64_t c, const string& d)
+  {
+    ostringstream os;
+    if (!n.empty ())
+      os << n << ':';
+    os << l << ':' << c << ": error: " << d;
+    return os.str ();
+  }
+
+  tab_parsing::
+  tab_parsing (const string& n, uint64_t l, uint64_t c, const string& d)
+      : runtime_error (format (n, l, c, d)),
+        name (n), line (l), column (c), description (d)
+  {
+  }
+}