version0.7.10

1 files changed, 166 insertions, 0 deletions

diff --git a/lib/pure/lexbase.nim b/lib/pure/lexbase.nim
new file mode 100644
index 000000000..bb207e92a
--- /dev/null
+++ b/lib/pure/lexbase.nim
@@ -0,0 +1,166 @@
+#
+#
+#           The Nimrod Compiler
+#        (c) Copyright 2009 Andreas Rumpf
+#
+#    See the file "copying.txt", included in this
+#    distribution, for details about the copyright.
+#
+
+## This module implements a base object of a lexer with efficient buffer
+## handling. Only at line endings checks are necessary if the buffer
+## needs refilling.
+
+import
+  strutils, streams
+
+const
+  EndOfFile* = '\0'           ## end of file marker
+  NewLines* = {'\c', '\L'}
+
+# Buffer handling:
+#  buf:
+#  "Example Text\n ha!"   bufLen = 17
+#   ^pos = 0     ^ sentinel = 12
+#
+
+type
+  TBaseLexer* = object of TObject ## the base lexer. Inherit your lexer from
+                                  ## this object.
+    bufpos*: int              ## the current position within the buffer
+    buf*: cstring             ## the buffer itself
+    bufLen*: int              ## length of buffer in characters
+    input: PStream            ## the input stream
+    LineNumber*: int          ## the current line number
+    sentinel: int
+    lineStart: int            # index of last line start in buffer
+    fileOpened: bool
+
+proc open*(L: var TBaseLexer, input: PStream, bufLen: int = 8192)
+  ## inits the TBaseLexer with a stream to read from
+
+proc close*(L: var TBaseLexer)
+  ## closes the base lexer. This closes `L`'s associated stream too.
+
+proc getCurrentLine*(L: TBaseLexer, marker: bool = true): string
+  ## retrieves the current line.
+
+proc getColNumber*(L: TBaseLexer, pos: int): int
+  ## retrieves the current column.
+
+proc HandleCR*(L: var TBaseLexer, pos: int): int
+  ## Call this if you scanned over '\c' in the buffer; it returns the the
+  ## position to continue the scanning from. `pos` must be the position
+  ## of the '\c'.
+proc HandleLF*(L: var TBaseLexer, pos: int): int
+  ## Call this if you scanned over '\L' in the buffer; it returns the the
+  ## position to continue the scanning from. `pos` must be the position
+  ## of the '\L'.
+
+# implementation
+
+const
+  chrSize = sizeof(char)
+
+proc close(L: var TBaseLexer) =
+  dealloc(L.buf)
+  L.input.close(L.input)
+
+proc FillBuffer(L: var TBaseLexer) =
+  var
+    charsRead, toCopy, s: int # all are in characters,
+                              # not bytes (in case this
+                              # is not the same)
+    oldBufLen: int
+  # we know here that pos == L.sentinel, but not if this proc
+  # is called the first time by initBaseLexer()
+  assert(L.sentinel < L.bufLen)
+  toCopy = L.BufLen - L.sentinel - 1
+  assert(toCopy >= 0)
+  if toCopy > 0:
+    MoveMem(L.buf, addr(L.buf[L.sentinel + 1]), toCopy * chrSize) # "moveMem" handles overlapping regions
+  charsRead = L.input.readData(L.input, addr(L.buf[toCopy]),
+                               (L.sentinel + 1) * chrSize) div chrSize
+  s = toCopy + charsRead
+  if charsRead < L.sentinel + 1:
+    L.buf[s] = EndOfFile      # set end marker
+    L.sentinel = s
+  else:
+    # compute sentinel:
+    dec(s)                    # BUGFIX (valgrind)
+    while true:
+      assert(s < L.bufLen)
+      while (s >= 0) and not (L.buf[s] in NewLines): Dec(s)
+      if s >= 0:
+        # we found an appropriate character for a sentinel:
+        L.sentinel = s
+        break
+      else:
+        # rather than to give up here because the line is too long,
+        # double the buffer's size and try again:
+        oldBufLen = L.BufLen
+        L.bufLen = L.BufLen * 2
+        L.buf = cast[cstring](realloc(L.buf, L.bufLen * chrSize))
+        assert(L.bufLen - oldBuflen == oldBufLen)
+        charsRead = L.input.ReadData(L.input, addr(L.buf[oldBufLen]),
+                                     oldBufLen * chrSize) div chrSize
+        if charsRead < oldBufLen:
+          L.buf[oldBufLen + charsRead] = EndOfFile
+          L.sentinel = oldBufLen + charsRead
+          break
+        s = L.bufLen - 1
+
+proc fillBaseLexer(L: var TBaseLexer, pos: int): int =
+  assert(pos <= L.sentinel)
+  if pos < L.sentinel:
+    result = pos + 1          # nothing to do
+  else:
+    fillBuffer(L)
+    L.bufpos = 0              # XXX: is this really correct?
+    result = 0
+  L.lineStart = result
+
+proc HandleCR(L: var TBaseLexer, pos: int): int =
+  assert(L.buf[pos] == '\c')
+  inc(L.linenumber)
+  result = fillBaseLexer(L, pos)
+  if L.buf[result] == '\L':
+    result = fillBaseLexer(L, result)
+
+proc HandleLF(L: var TBaseLexer, pos: int): int =
+  assert(L.buf[pos] == '\L')
+  inc(L.linenumber)
+  result = fillBaseLexer(L, pos) #L.lastNL := result-1; // BUGFIX: was: result;
+
+proc skip_UTF_8_BOM(L: var TBaseLexer) =
+  if (L.buf[0] == '\xEF') and (L.buf[1] == '\xBB') and (L.buf[2] == '\xBF'):
+    inc(L.bufpos, 3)
+    inc(L.lineStart, 3)
+
+proc open(L: var TBaseLexer, input: PStream, bufLen: int = 8192) =
+  assert(bufLen > 0)
+  assert(input != nil)
+  L.input = input
+  L.bufpos = 0
+  L.bufLen = bufLen
+  L.buf = cast[cstring](alloc(bufLen * chrSize))
+  L.sentinel = bufLen - 1
+  L.lineStart = 0
+  L.linenumber = 1            # lines start at 1
+  fillBuffer(L)
+  skip_UTF_8_BOM(L)
+
+proc getColNumber(L: TBaseLexer, pos: int): int =
+  result = abs(pos - L.lineStart)
+
+proc getCurrentLine(L: TBaseLexer, marker: bool = true): string =
+  var i: int
+  result = ""
+  i = L.lineStart
+  while not (L.buf[i] in {'\c', '\L', EndOfFile}):
+    add(result, L.buf[i])
+    inc(i)
+  add(result, "\n")
+  if marker:
+    add(result, RepeatChar(getColNumber(L, L.bufpos)) & "^\n")
+