# $Id: tableparser.py 7320 2012-01-19 22:33:02Z milde $ # Author: David Goodger # Copyright: This module has been placed in the public domain. """ This module defines table parser classes,which parse plaintext-graphic tables and produce a well-formed data structure suitable for building a CALS table. :Classes: - `GridTableParser`: Parse fully-formed tables represented with a grid. - `SimpleTableParser`: Parse simple tables, delimited by top & bottom borders. :Exception class: `TableMarkupError` :Function: `update_dict_of_lists()`: Merge two dictionaries containing list values. """ __docformat__ = 'reStructuredText' import re import sys from docutils import DataError from docutils.utils import strip_combining_chars class TableMarkupError(DataError): """ Raise if there is any problem with table markup. The keyword argument `offset` denotes the offset of the problem from the table's start line. """ def __init__(self, *args, **kwargs): self.offset = kwargs.pop('offset', 0) DataError.__init__(self, *args) class TableParser: """ Abstract superclass for the common parts of the syntax-specific parsers. """ head_body_separator_pat = None """Matches the row separator between head rows and body rows.""" double_width_pad_char = '\x00' """Padding character for East Asian double-width text.""" def parse(self, block): """ Analyze the text `block` and return a table data structure. Given a plaintext-graphic table in `block` (list of lines of text; no whitespace padding), parse the table, construct and return the data necessary to construct a CALS table or equivalent. Raise `TableMarkupError` if there is any problem with the markup. """ self.setup(block) self.find_head_body_sep() self.parse_table() structure = self.structure_from_cells() return structure def find_head_body_sep(self): """Look for a head/body row separator line; store the line index.""" for i in range(len(self.block)): line = self.block[i] if self.head_body_separator_pat.match(line): if self.head_body_sep: raise TableMarkupError( 'Multiple head/body row separators ' '(table lines %s and %s); only one allowed.' % (self.head_body_sep+1, i+1), offset=i) else: self.head_body_sep = i self.block[i] = line.replace('=', '-') if self.head_body_sep == 0 or self.head_body_sep == (len(self.block) - 1): raise TableMarkupError('The head/body row separator may not be ' 'the first or last line of the table.', offset=i) class GridTableParser(TableParser): """ Parse a grid table using `parse()`. Here's an example of a grid table:: +------------------------+------------+----------+----------+ | Header row, column 1 | Header 2 | Header 3 | Header 4 | +========================+============+==========+==========+ | body row 1, column 1 | column 2 | column 3 | column 4 | +------------------------+------------+----------+----------+ | body row 2 | Cells may span columns. | +------------------------+------------+---------------------+ | body row 3 | Cells may | - Table cells | +------------------------+ span rows. | - contain | | body row 4 | | - body elements. | +------------------------+------------+---------------------+ Intersections use '+', row separators use '-' (except for one optional head/body row separator, which uses '='), and column separators use '|'. Passing the above table to the `parse()` method will result in the following data structure:: ([24, 12, 10, 10], [[(0, 0, 1, ['Header row, column 1']), (0, 0, 1, ['Header 2']), (0, 0, 1, ['Header 3']), (0, 0, 1, ['Header 4'])]], [[(0, 0, 3, ['body row 1, column 1']), (0, 0, 3, ['column 2']), (0, 0, 3, ['column 3']), (0, 0, 3, ['column 4'])], [(0, 0, 5, ['body row 2']), (0, 2, 5, ['Cells may span columns.']), None, None], [(0, 0, 7, ['body row 3']), (1, 0, 7, ['Cells may', 'span rows.', '']), (1, 1, 7, ['- Table cells', '- contain', '- body elements.']), None], [(0, 0, 9, ['body row 4']), None, None, None]]) The first item is a list containing column widths (colspecs). The second item is a list of head rows, and the third is a list of body rows. Each row contains a list of cells. Each cell is either None (for a cell unused because of another cell's span), or a tuple. A cell tuple contains four items: the number of extra rows used by the cell in a vertical span (morerows); the number of extra columns used by the cell in a horizontal span (morecols); the line offset of the first line of the cell contents; and the cell contents, a list of lines of text. """ head_body_separator_pat = re.compile(r'\+=[=+]+=\+ *$') def setup(self, block): self.block = block[:] # make a copy; it may be modified self.block.disconnect() # don't propagate changes to parent self.bottom = len(block) - 1 self.right = len(block[0]) - 1 self.head_body_sep = None self.done = [-1] * len(block[0]) self.cells = [] self.rowseps = {0: [0]} self.colseps = {0: [0]} def parse_table(self): """ Start with a queue of upper-left corners, containing the upper-left corner of the table itself. Trace out one rectangular cell, remember it, and add its upper-right and lower-left corners to the queue of potential upper-left corners of further cells. Process the queue in top-to-bottom order, keeping track of how much of each text column has been seen. We'll end up knowing all the row and column boundaries, cell positions and their dimensions. """ corners = [(0, 0)] while corners: top, left = corners.pop(0) if top == self.bottom or left == self.right \ or top <= self.done[left]: continue result = self.scan_cell(top, left) if not result: continue bottom, right, rowseps, colseps = result update_dict_of_lists(self.rowseps, rowseps) update_dict_of_lists(self.colseps, colseps) self.mark_done(top, left, bottom, right) cellblock = self.block.get_2D_block(top + 1, left + 1, bottom, right) cellblock.disconnect() # lines in cell can't sync with parent cellblock.replace(self.double_width_pad_char, '') self.cells.append((top, left, bottom, right, cellblock)) corners.extend([(top, right), (bottom, left)]) corners.sort() if not self.check_parse_complete(): raise TableMarkupError('Malformed table; parse incomplete.') def mark_done(self, top, left, bottom, right): """For keeping track of how much of each text column has been seen.""" before = top - 1 after = bottom - 1 for col in range(left, right): assert self.done[col] == before self.done[col] = after def check_parse_complete(self): """Each text column should have been completely seen.""" last = self.bottom - 1 for col in range(self.right): if self.done[col] != last: return False return True def scan_cell(self, top, left): """Starting at the top-left corner, start tracing out a cell.""" assert self.block[top][left] == '+' result = self.scan_right(top, left) return result def scan_right(self, top, left): """ Look for the top-right corner of the cell, and make note of all column boundaries ('+'). """ colseps = {} line = self.block[top] for i in range(left + 1, self.right + 1): if line[i] == '+': colseps[i] = [top] result = self.scan_down(top, left, i) if result: bottom, rowseps, newcolseps = result update_dict_of_lists(colseps, newcolseps) return bottom, i, rowseps, colseps elif line[i] != '-': return None return None def scan_down(self, top, left, right): """ Look for the bottom-right corner of the cell, making note of all row boundaries. """ rowseps = {} for i in range(top + 1, self.bottom + 1): if self.block[i][right] == '+': rowseps[i] = [right] result = self.scan_left(top, left, i, right) if result: newrowseps, colseps = result update_dict_of_lists(rowseps, newrowseps) return i, rowseps, colseps elif self.block[i][right] != '|': return None return None def scan_left(self, top, left, bottom, right): """ Noting column boundaries, look for the bottom-left corner of the cell. It must line up with the starting point. """ colseps = {} line = self.block[bottom] for i in range(right - 1, left, -1): if line[i] == '+': colseps[i] = [bottom] elif line[i] != '-': return None if line[left] != '+': return None result = self.scan_up(top, left, bottom, right) if result is not None: rowseps = result return rowseps, colseps return None def scan_up(self, top, left, bottom, right): """ Noting row boundaries, see if we can return to the starting point. """ rowseps = {} for i in range(bottom - 1, top, -1): if self.block[i][left] == '+': rowseps[i] = [left] elif self.block[i][left] != '|': return None return rowseps def structure_from_cells(self): """ From the data collected by `scan_cell()`, convert to the final data structure. """ rowseps = self.rowseps.keys() # list of row boundaries rowseps.sort() rowindex = {} for i in range(len(rowseps)): rowindex[rowseps[i]] = i # row boundary -> row number mapping colseps = self.colseps.keys() # list of column boundaries colseps.sort() colindex = {} for i in range(len(colseps)): colindex[colseps[i]] = i # column boundary -> col number map colspecs = [(colseps[i] - colseps[i - 1] - 1) for i in range(1, len(colseps))] # list of column widths # prepare an empty table with the correct number of rows & columns onerow = [None for i in range(len(colseps) - 1)] rows = [onerow[:] for i in range(len(rowseps) - 1)] # keep track of # of cells remaining; should reduce to zero remaining = (len(rowseps) - 1) * (len(colseps) - 1) for top, left, bottom, right, block in self.cells: rownum = rowindex[top] colnum = colindex[left] assert rows[rownum][colnum] is None, ( 'Cell (row %s, column %s) already used.' % (rownum + 1, colnum + 1)) morerows = rowindex[bottom] - rownum - 1 morecols = colindex[right] - colnum - 1 remaining -= (morerows + 1) * (morecols + 1) # write the cell into the table rows[rownum][colnum] = (morerows, morecols, top + 1, block) assert remaining == 0, 'Unused cells remaining.' if self.head_body_sep: # separate head rows from body rows numheadrows = rowindex[self.head_body_sep] headrows = rows[:numheadrows] bodyrows = rows[numheadrows:] else: headrows = [] bodyrows = rows return (colspecs, headrows, bodyrows) class SimpleTableParser(TableParser): """ Parse a simple table using `parse()`. Here's an example of a simple table:: ===== ===== col 1 col 2 ===== ===== 1 Second column of row 1. 2 Second column of row 2. Second line of paragraph. 3 - Second column of row 3. - Second item in bullet list (row 3, column 2). 4 is a span ------------ 5 ===== ===== Top and bottom borders use '=', column span underlines use '-', column separation is indicated with spaces. Passing the above table to the `parse()` method will result in the following data structure, whose interpretation is the same as for `GridTableParser`:: ([5, 25], [[(0, 0, 1, ['col 1']), (0, 0, 1, ['col 2'])]], [[(0, 0, 3, ['1']), (0, 0, 3, ['Second column of row 1.'])], [(0, 0, 4, ['2']), (0, 0, 4, ['Second column of row 2.', 'Second line of paragraph.'])], [(0, 0, 6, ['3']), (0, 0, 6, ['- Second column of row 3.', '', '- Second item in bullet', ' list (row 3, column 2).'])], [(0, 1, 10, ['4 is a span'])], [(0, 0, 12, ['5']), (0, 0, 12, [''])]]) """ head_body_separator_pat = re.compile('=[ =]*$') span_pat = re.compile('-[ -]*$') def setup(self, block): self.block = block[:] # make a copy; it will be modified self.block.disconnect() # don't propagate changes to parent # Convert top & bottom borders to column span underlines: self.block[0] = self.block[0].replace('=', '-') self.block[-1] = self.block[-1].replace('=', '-') self.head_body_sep = None self.columns = [] self.border_end = None self.table = [] self.done = [-1] * len(block[0]) self.rowseps = {0: [0]} self.colseps = {0: [0]} def parse_table(self): """ First determine the column boundaries from the top border, then process rows. Each row may consist of multiple lines; accumulate lines until a row is complete. Call `self.parse_row` to finish the job. """ # Top border must fully describe all table columns. self.columns = self.parse_columns(self.block[0], 0) self.border_end = self.columns[-1][1] firststart, firstend = self.columns[0] offset = 1 # skip top border start = 1 text_found = None while offset < len(self.block): line = self.block[offset] if self.span_pat.match(line): # Column span underline or border; row is complete. self.parse_row(self.block[start:offset], start, (line.rstrip(), offset)) start = offset + 1 text_found = None elif line[firststart:firstend].strip(): # First column not blank, therefore it's a new row. if text_found and offset != start: self.parse_row(self.block[start:offset], start) start = offset text_found = 1 elif not text_found: start = offset + 1 offset += 1 def parse_columns(self, line, offset): """ Given a column span underline, return a list of (begin, end) pairs. """ cols = [] end = 0 while True: begin = line.find('-', end) end = line.find(' ', begin) if begin < 0: break if end < 0: end = len(line) cols.append((begin, end)) if self.columns: if cols[-1][1] != self.border_end: raise TableMarkupError('Column span incomplete in table ' 'line %s.' % (offset+1), offset=offset) # Allow for an unbounded rightmost column: cols[-1] = (cols[-1][0], self.columns[-1][1]) return cols def init_row(self, colspec, offset): i = 0 cells = [] for start, end in colspec: morecols = 0 try: assert start == self.columns[i][0] while end != self.columns[i][1]: i += 1 morecols += 1 except (AssertionError, IndexError): raise TableMarkupError('Column span alignment problem ' 'in table line %s.' % (offset+2), offset=offset+1) cells.append([0, morecols, offset, []]) i += 1 return cells def parse_row(self, lines, start, spanline=None): """ Given the text `lines` of a row, parse it and append to `self.table`. The row is parsed according to the current column spec (either `spanline` if provided or `self.columns`). For each column, extract text from each line, and check for text in column margins. Finally, adjust for insignificant whitespace. """ if not (lines or spanline): # No new row, just blank lines. return if spanline: columns = self.parse_columns(*spanline) span_offset = spanline[1] else: columns = self.columns[:] span_offset = start self.check_columns(lines, start, columns) row = self.init_row(columns, start) for i in range(len(columns)): start, end = columns[i] cellblock = lines.get_2D_block(0, start, len(lines), end) cellblock.disconnect() # lines in cell can't sync with parent cellblock.replace(self.double_width_pad_char, '') row[i][3] = cellblock self.table.append(row) def check_columns(self, lines, first_line, columns): """ Check for text in column margins and text overflow in the last column. Raise TableMarkupError if anything but whitespace is in column margins. Adjust the end value for the last column if there is text overflow. """ # "Infinite" value for a dummy last column's beginning, used to # check for text overflow: columns.append((sys.maxint, None)) lastcol = len(columns) - 2 # combining characters do not contribute to the column width lines = [strip_combining_chars(line) for line in lines] for i in range(len(columns) - 1): start, end = columns[i] nextstart = columns[i+1][0] offset = 0 for line in lines: if i == lastcol and line[end:].strip(): text = line[start:].rstrip() new_end = start + len(text) columns[i] = (start, new_end) main_start, main_end = self.columns[-1] if new_end > main_end: self.columns[-1] = (main_start, new_end) elif line[end:nextstart].strip(): raise TableMarkupError('Text in column margin ' 'in table line %s.' % (first_line+offset+1), offset=first_line+offset) offset += 1 columns.pop() def structure_from_cells(self): colspecs = [end - start for start, end in self.columns] first_body_row = 0 if self.head_body_sep: for i in range(len(self.table)): if self.table[i][0][2] > self.head_body_sep: first_body_row = i break return (colspecs, self.table[:first_body_row], self.table[first_body_row:]) def update_dict_of_lists(master, newdata): """ Extend the list values of `master` with those from `newdata`. Both parameters must be dictionaries containing list values. """ for key, values in newdata.items(): master.setdefault(key, []).extend(values)