libreoffice python 操作word及excel文档的方法

yipeiwu_com5年前Python基础

1、开始、关闭libreoffice服务;

开始之前同步字体文件时间,是因为创建soffice服务时,服务会检查所需加载的文件的时间,如果其认为时间不符,则其可能会重新加载,耗时较长,因此需事先统一时间。

使用时如果需要多次调用,最后每次调用均开启后关闭,否则libreoffice会创建一个缓存文档并越用越大,处理时间会增加。

class OfficeProcess(object):
  def __init__(self):
    self.p = 0
    subprocess.Popen('find /usr/share/fonts | xargs touch -m -t 201801010000.00', shell=True)

  def start_office(self):
    self.p = subprocess.Popen('soffice --pidfile=sof.pid --invisible --accept="socket,host=localhost,port=2002;urp;"', shell=True)
    while True:
      try:
        local_context = uno.getComponentContext()
        resolver = local_context.getServiceManager().createInstanceWithContext('com.sun.star.bridge.UnoUrlResolver', local_context)
        resolver.resolve('uno:socket,host=localhost,port=2002;urp;StarOffice.ComponentContext')
        return
      except:
        print(ts(), "wait for connecting soffice...")
        time.sleep(1)
        continue

  def stop_office(self):
    with open("sof.pid", "rb") as f:
      try:
        os.kill(int(f.read()), signal.SIGTERM)
        self.p.wait()
      except:
        pass

2、init service manager

local_context = uno.getComponentContext()
    service_manager = local_context.getServiceManager()
    resolver = service_manager.createInstanceWithContext('com.sun.star.bridge.UnoUrlResolver', local_context)
    self.ctx = resolver.resolve('uno:socket,host=localhost,port=2002;urp;StarOffice.ComponentContext')
    self.smgr = self.ctx.ServiceManager
    self.desktop = self.smgr.createInstanceWithContext('com.sun.star.frame.Desktop', self.ctx)

3、从二进制数据中读取doc文档

def ImportFromMemory(self, data):
    istream = self.smgr.createInstanceWithContext('com.sun.star.io.SequenceInputStream', self.ctx)
    istream.initialize((uno.ByteSequence(data), ))
    pv = PropertyValue()
    pv.Name = 'InputStream'
    pv.Value = istream
    self.doc = {'doc': []}
    try:
      self.document = self.desktop.loadComponentFromURL('private:stream/swriter', '_blank', 0, (pv, ))
      self.text = self.document.getText()
    except:
      self.text = None

4、读取doc文档中的数据

def ExportToJson(self):
    try:
      l = self.__ParseText(self.text, self.__Callback(self.doc['doc']))
      self.doc['length'] = l
    except:
      self.doc = {'doc': [], 'length': 0}
    return json.dumps(self.doc)

@staticmethod
  def __Callback(alist):
    def Append(sth):
      alist.append(sth)
    return Append

def __ParseText(self, text, func):
    l = 0
    text_it = text.createEnumeration()
    while text_it.hasMoreElements():
      element = text_it.nextElement()
      if element.supportsService('com.sun.star.text.Paragraph'):
        l += self.__ParseParagraph(element, func)
      elif element.supportsService('com.sun.star.text.TextTable'):
        l += self.__ParseTable(element, func)
      else:
        pass
    return l
def __ParseParagraph(self, paragraph, func):
    p = {'paragraph': []}
    l = 0
    paragraph_it = paragraph.createEnumeration()
    while paragraph_it.hasMoreElements():
      portion = paragraph_it.nextElement()
      if portion.TextPortionType == 'Text':
        l += self.__ParsePortionText(portion, self.__Callback(p['paragraph']))
      elif portion.TextPortionType == 'SoftPageBreak':
        pass
      elif portion.TextPortionType == 'TextField':
        l += self.__ParsePortionText(portion, self.__Callback(p['paragraph']))
      else:
        l += self.__ParseTextContent(portion, self.__Callback(p['paragraph']))
    if hasattr(paragraph, 'createContentEnumeration'):
      l += self.__ParseTextContent(paragraph, self.__Callback(p['paragraph']))
    p['length'] = l
    func(p)
    return l

  def __ParseTextContent(self, textcontent, func):
    l = 0
    content_it = textcontent.createContentEnumeration('com.sun.star.text.TextContent')
    while content_it.hasMoreElements():
      element = content_it.nextElement()
      if element.supportsService('com.sun.star.text.TextGraphicObject'):
        l += self.__ParsePortionGraphic(element, func)
      elif element.supportsService('com.sun.star.text.TextEmbeddedObject'):
        pass
      elif element.supportsService('com.sun.star.text.TextFrame'):
        l += self.__ParseFrame(element, func)
      elif element.supportsService('com.sun.star.drawing.GroupShape'):
        l += self.__ParseGroup(element, func)
      else:
        pass
    return l

  def __ParseFrame(self, frame, func):
    f = {'frame': []}
    l = self.__ParseText(frame.getText(), self.__Callback(f['frame']))
    f['length'] = l
    func(f)
    return l

  def __ParseGroup(self, group, func):
    l = 0
    for i in range(group.getCount()):
      it = group.getByIndex(i)
      if it.supportsService('com.sun.star.drawing.Text'):
        l += self.__ParseFrame(it, func)
      else:
        pass
    return l

  def __ParsePortionText(self, portion_text, func):
    func({'portion': portion_text.String, 'length': len(portion_text.String)})
    return len(portion_text.String)

  def __ParsePortionGraphic(self, portion_graphic, func):
    gp = self.smgr.createInstanceWithContext('com.sun.star.graphic.GraphicProvider', self.ctx)
    stream = self.smgr.createInstanceWithContext('com.sun.star.io.TempFile', self.ctx)
    pv1 = PropertyValue()
    pv1.Name = 'OutputStream'
    pv1.Value = stream
    pv2 = PropertyValue()
    pv2.Name = 'MimeType'
    pv2.Value = 'image/png'
    gp.storeGraphic(portion_graphic.Graphic, (pv1, pv2))
    stream.getOutputStream().flush()
    stream.seek(0)
    l = stream.getInputStream().available()
    b = uno.ByteSequence(b'')
    stream.seek(0)
    l, b = stream.getInputStream().readBytes(b, l)
    img = {'image': base64.b64encode(b.value).decode('ascii')}
    img['height'] = portion_graphic.Height
    img['width'] = portion_graphic.Width
    img['actualheight'] = portion_graphic.ActualSize.Height
    img['actualwidth'] = portion_graphic.ActualSize.Width
    img['croptop'] = portion_graphic.GraphicCrop.Top
    img['cropbottom'] = portion_graphic.GraphicCrop.Bottom
    img['cropleft'] = portion_graphic.GraphicCrop.Left
    img['cropright'] = portion_graphic.GraphicCrop.Right
    img['length'] = 0
    func(img)
    return 0

  def __ParseTable(self, table, func):
    l = 0
    try:
      matrix = self.__GetTableMatrix(table)
      seps = self.__GetTableSeparators(table)
      t = {}
      count = 0
      for ri in matrix.keys():
        t[ri] = {}
        for ci in matrix[ri].keys():
          t[ri][ci] = dict(matrix[ri][ci])
          del t[ri][ci]['cell']
          t[ri][ci]['content'] = []
          l += self.__ParseText(matrix[ri][ci]['cell'], self.__Callback(t[ri][ci]['content']))
          count += t[ri][ci]['rowspan'] * t[ri][ci]['colspan']
      if count != len(t) * len(seps):
        raise ValueError('count of cells error')
      func({'table': t, 'row': len(t), 'column': len(seps), 'length': l, 'tableid': self.table_id})
      self.table_id += 1
    except:
      l = 0
      print('discard wrong table')
    return l

  @staticmethod
  def __GetTableSeparators(table):
    result = [table.TableColumnRelativeSum]
    for ri in range(table.getRows().getCount()):
      result += [s.Position for s in table.getRows().getByIndex(ri).TableColumnSeparators]
    result = sorted(set(result))
    for i in range(len(result) - 1):
      result[i] += 1 if result[i] + 1 == result[i + 1] else 0
    return sorted(set(result))

  @staticmethod
  def __NameToRC(name):
    r = int(re.sub('[A-Za-z]', '', name)) - 1
    cstr = re.sub('[0-9]', '', name)
    c = 0
    for i in range(len(cstr)):
      if cstr[i] >= 'A' and cstr[i] <= 'Z':
        c = c * 52 + ord(cstr[i]) - ord('A')
      else:
        c = c * 52 + 26 + ord(cstr[i]) - ord('a')
    return r, c

  @staticmethod
  def __GetTableMatrix(table):
    result = {}
    for name in table.getCellNames():
      ri, ci = WordToJson.__NameToRC(name)
      cell = table.getCellByName(name)
      if ri not in result:
        result[ri] = {}
      result[ri][ci] = {'cell': cell, 'rowspan': cell.RowSpan, 'name': name}

    seps = WordToJson.__GetTableSeparators(table)
    for ri in result.keys():
      sep = [s.Position for s in table.getRows().getByIndex(ri).TableColumnSeparators] + [table.TableColumnRelativeSum]
      sep = sorted(set(sep))
      for ci in result[ri].keys():
        right = seps.index(sep[ci]) if sep[ci] in seps else seps.index(sep[ci] + 1)
        left = -1 if ci == 0 else seps.index(sep[ci - 1]) if sep[ci - 1] in seps else seps.index(sep[ci - 1] + 1)
        result[ri][ci]['colspan'] = right - left
    return result

5、写doc文档

self.doco = self.desktop.loadComponentFromURL('private:factory/swriter', '_blank', 0, ())
    self.texto = self.doco.getText()
    self.cursoro = self.texto.createTextCursor()
    self.cursoro.ParaBottomMargin = 500
def __WriteText(self, text, texto, cursoro):
    for it in text:
      if 'paragraph' in it:
        self.__WriteParagraph(it, texto, cursoro)
      elif 'image' in it:
        self.__WritePortionGraphic(it, texto, cursoro)
      elif 'table' in it:
        self.__WriteTable(it, texto, cursoro)

  def __WriteParagraph(self, paragraph, texto, cursoro):
    if paragraph['length'] > 0:
      if 'result' in paragraph:
        for it in paragraph['result']:
          texto.insertString(cursoro, it['trans_sen'], False)
      else:
        texto.insertString(cursoro, paragraph['paragraph'], False)
      texto.insertControlCharacter(cursoro, ControlCharacter.PARAGRAPH_BREAK, False)

  def __WritePortionGraphic(self, portion_graphic, texto, cursoro):
    png_base64 = portion_graphic['image']
    png = base64.b64decode(png_base64)
    gp = self.smgr.createInstanceWithContext('com.sun.star.graphic.GraphicProvider', self.ctx)
    istream = self.smgr.createInstanceWithContext('com.sun.star.io.SequenceInputStream', self.ctx)
    istream.initialize((uno.ByteSequence(png), ))
    pv = PropertyValue()
    pv.Name = 'InputStream'
    pv.Value = istream

    actualsize = uno.createUnoStruct('com.sun.star.awt.Size')
    actualsize.Height = portion_graphic['actualheight'] if 'actualheight' in portion_graphic else portion_graphic['height']
    actualsize.Width = portion_graphic['actualwidth'] if 'actualwidth' in portion_graphic else portion_graphic['width']
    graphiccrop = uno.createUnoStruct('com.sun.star.text.GraphicCrop')
    graphiccrop.Top = portion_graphic['croptop'] if 'croptop' in portion_graphic else 0
    graphiccrop.Bottom = portion_graphic['cropbottom'] if 'cropbottom' in portion_graphic else 0
    graphiccrop.Left = portion_graphic['cropleft'] if 'cropleft' in portion_graphic else 0
    graphiccrop.Right = portion_graphic['cropright'] if 'cropright' in portion_graphic else 0

    image = self.doco.createInstance('com.sun.star.text.TextGraphicObject')
    image.Surround = NONE
    image.Graphic = gp.queryGraphic((pv, ))
    image.Height = portion_graphic['height']
    image.Width = portion_graphic['width']
    image.setPropertyValue('ActualSize', actualsize)
    image.setPropertyValue('GraphicCrop', graphiccrop)
    texto.insertTextContent(cursoro, image, False)
    texto.insertControlCharacter(cursoro, ControlCharacter.PARAGRAPH_BREAK, False)

  def __WriteTable(self, table, texto, cursoro):
    tableo = self.doco.createInstance('com.sun.star.text.TextTable')
    tableo.initialize(table['row'], table['column'])
    texto.insertTextContent(cursoro, tableo, False)
#    texto.insertControlCharacter(cursoro, ControlCharacter.PARAGRAPH_BREAK, False)
    tcursoro = tableo.createCursorByCellName("A1")

    hitbug = False
    if table['row'] > 1:
      tcursoro.goDown(1, True)
      hitbug = tcursoro.getRangeName() == 'A1'

    for ri in sorted([int(r) for r in table['table'].keys()]):
      rs = table['table'][str(ri)]
      for ci in sorted([int(c) for c in rs.keys()]):
        cell = rs[str(ci)]
        if hitbug == False and (cell['rowspan'] > 1 or cell['colspan'] > 1):
          tcursoro.gotoCellByName(cell['name'], False)
          if cell['rowspan'] > 1:
            tcursoro.goDown(cell['rowspan'] - 1, True)
          if cell['colspan'] > 1:
            tcursoro.goRight(cell['colspan'] - 1, True)
          tcursoro.mergeRange()
        ctexto = tableo.getCellByName(cell['name'])
        if ctexto == None:
          continue
        ccursoro = ctexto.createTextCursor()
        ccursoro.CharWeight = FontWeight.NORMAL
        ccursoro.CharWeightAsian = FontWeight.NORMAL
        ccursoro.ParaAdjust = LEFT
        self.__WriteText(cell['content'], ctexto, ccursoro)

6、生成二进制的doc文档数据

    streamo = self.smgr.createInstanceWithContext('com.sun.star.io.Pipe', self.ctx)
    self.doco.storeToURL('private:stream', (PropertyValue('FilterName', 0, 'MS Word 2007 XML', 0), PropertyValue('OutputStream', 0, streamo, 0)))
    streamo.flush()
    _, datao = streamo.readBytes(None, streamo.available())

7、从doc文档数据生成pdf的二进制数据

    streamo = self.smgr.createInstanceWithContext('com.sun.star.io.Pipe', self.ctx)
    self.doco.storeToURL('private:stream', (PropertyValue('FilterName', 0, 'writer_pdf_Export', 0), PropertyValue('OutputStream', 0, streamo, 0)))
    streamo.flush()
    _, datap = streamo.readBytes(None, streamo.available())

8、读取excel二进制数据

  def ImportFromMemory(self, data):
    istream = self.smgr.createInstanceWithContext('com.sun.star.io.SequenceInputStream', self.ctx)
    istream.initialize((uno.ByteSequence(data), ))
    pv = PropertyValue()
    pv.Name = 'InputStream'
    pv.Value = istream
    self.doc = {'doc': []}
    try:
      print("before loadComponentFromURL")
      self.document = self.desktop.loadComponentFromURL('private:stream/scalc', '_blank', 0, (pv, ))
      self.sheets = self.document.getSheets()
      print("ImportFromMemory done")
    except:
      print("ImportFromMemory failed")
      self.sheets = None

9、读取excel的文本数据

  def ExportToJson(self):
    try:
      l = self.__ParseText(self.sheets, self.__Callback(self.doc['doc']))
      self.doc['length'] = l
    except:
      self.doc = {'doc': [], 'length': 0}
    return json.dumps(self.doc)
  def __ParseText(self, sheets, func):
    l = 0
    sheets_it = sheets.createEnumeration()
    while sheets_it.hasMoreElements():
      element = sheets_it.nextElement()
      if element.supportsService('com.sun.star.sheet.Spreadsheet'):
        l += self.__ParseSpreadsheet(element, func)
    return l

  def __ParseSpreadsheet(self, spreadsheet, func):
    l = 0
    p = {'spreadsheet': []}
    visible_cells_it = spreadsheet.queryVisibleCells().getCells().createEnumeration()
    while visible_cells_it.hasMoreElements():
      cell = visible_cells_it.nextElement()
      type = cell.getType()
      if type == self.EMPTY:
        print("cell.type==empty")
      elif type == self.VALUE:
        print("cell.type==VALUE", "value=", cell.getValue(), cell.getCellAddress ())
      elif type == self.TEXT:
        print("cell.type==TEXT","content=", cell.getString().encode("UTF-8"), cell.getCellAddress ())
        l += self.__ParseCellText(spreadsheet, cell, self.__Callback(p['spreadsheet']))
        print("__ParseCellText=", p)
      elif type == self.FORMULA:
        print("cell.type==FORMULA", "formula=", cell.getValue())
    p['length'] = l
    func(p)
    return l

  def __ParseCellText(self, sheet, cell, func):
    try:
      x = cell.getCellAddress().Column
      y = cell.getCellAddress().Row
      sheetname = sheet.getName()
    except:
      x = -1
      y = -1
      sheetname = None
    func({'celltext': cell.getString(), 'x': x, 'y': y, 'sheetname': sheetname, 'length': len(cell.getString())})
    return len(cell.getString())
     self.EMPTY = uno.Enum("com.sun.star.table.CellContentType", "EMPTY")
    self.TEXT = uno.Enum("com.sun.star.table.CellContentType", "TEXT")
    self.FORMULA = uno.Enum("com.sun.star.table.CellContentType", "FORMULA")
    self.VALUE = uno.Enum("com.sun.star.table.CellContentType", "VALUE")

10、替换excel的文本信息

  def ImportFromJson(self, data):
    doc = json.loads(data)
    try:
      self.__WriteText(doc['doc'])
    except:
      pass
def __WriteText(self, text):
    print("__WriteText begin:", text)
    sheet = None
    for it in text:
      if 'paragraph' in it and 'sheetname' in it:
        if sheet == None or sheet.getName() != it['sheetname']:
          try:
            sheet = self.sheets.getByName(it['sheetname'])
            print("getsheet:", it['sheetname'], "=", sheet.getName())
          except:
            sheet = None
            continue
        self.__WriteParagraph(it, sheet)

  def __WriteParagraph(self, paragraph, sheet):
    print("__WriteParagraph")
    if paragraph['length'] > 0:
      try:
        x = paragraph['x']
        y = paragraph['y']
        print("getcell:", x, y)
        cell = sheet.getCellByPosition(x, y)
        print("getcell done")
      except:
        return
      if 'result' in paragraph:
        for it in paragraph['result']:
          print("cell=", cell.getString())
          cell.setString(it['trans_sen'])
          print("cell,", cell.getString(), ",done")

11、生成excel文档二进制数据

     streamo = self.smgr.createInstanceWithContext('com.sun.star.io.Pipe', self.ctx)
    self.document.storeToURL('private:stream', (PropertyValue('FilterName', 0, 'Calc MS Excel 2007 XML', 0), PropertyValue('OutputStream', 0, streamo, 0)))
    streamo.flush()
    _, datao = streamo.readBytes(None, streamo.available())

12、生成excel的pdf文档

    streamo = self.smgr.createInstanceWithContext('com.sun.star.io.Pipe', self.ctx)
    self.document.storeToURL('private:stream', (PropertyValue('FilterName', 0, 'calc_pdf_Export', 0), PropertyValue('OutputStream', 0, streamo, 0)))
    streamo.flush()
    _, datap = streamo.readBytes(None, streamo.available())

以上就是本文的全部内容,希望对大家的学习有所帮助,也希望大家多多支持【听图阁-专注于Python设计】。

相关文章

pytorch加载自定义网络权重的实现

在将自定义的网络权重加载到网络中时,报错: AttributeError: 'dict' object has no attribute 'seek'. You can only tor...

Django使用httpresponse返回用户头像实例代码

Django使用httpresponse返回用户头像实例代码

本文研究的主要是Django使用httpresponse返回用户头像,下面是相关实例代码。 当请求一个页面时,Django 把请求的 metadata 数据包装成一个 HttpReque...

django认证系统 Authentication使用详解

前言 Django自带一个用户认证系统,用于处理用户账户、群组、许可和基于cookie的用户会话。 Django的认证系统包含了身份验证和权限管理两部分。简单地说,身份验证用于核实某个用...

Numpy中矩阵matrix读取一列的方法及数组和矩阵的相互转换实例

Numpy matrix 必须是2维的,但是 numpy arrays (ndarrays) 可以是多维的(1D,2D,3D····ND),matrix是Array的一个小的分支,包含于...

django admin 后台实现三级联动的示例代码

在刚进公司的时候,要写一个需求,使用django的admin站点管理,实现一个二级联动的功能,因为要用到django自带的页面,因为不是自定义的,不能直接添加js代码。根据我自己的研究简...