public bool NotEmptyParagraphIsInADeclarationTable(Paragraph p)
{
if (p.InnerText == "")
{
return(false);
}
var body = WordDocument.MainDocumentPart.Document.Body;
if (p.Parent == body)
{
return(false);
}
var cell = p.Parent;
if (cell == null || cell.GetType() != typeof(TableCell))
{
return(false);
}
var row = cell.Parent;
if (row == null || row.GetType() != typeof(TableRow) || WordDocHolder.CountInnerVerticalBorders((TableRow)row) < 3)
{
return(false);
}
var table = row.Parent;
if (table != null && table.GetType().Name == "Table" && table.Descendants <TableRow>().Count() > 1)
{
return(true);
}
return(false);
}
public static int CountInnerVerticalBorders(TableRow row)
{
int cellIndex = 0;
HashSet <int> visibleBorders = new HashSet <int>();
var cells = row.Descendants <TableCell>().ToList();
foreach (var c in cells)
{
bool rightBorder = true;
bool leftBorder = true;
if (c?.TableCellProperties?.TableCellBorders != null)
{
var cellBorders = c.TableCellProperties.TableCellBorders;
if (!WordDocHolder.BorderIsVisible(cellBorders.LeftBorder))
{
leftBorder = false;
}
if (!WordDocHolder.BorderIsVisible(cellBorders.RightBorder))
{
rightBorder = false;
}
}
if (cellIndex > 0 && leftBorder)
{
visibleBorders.Add(cellIndex);
}
if (cellIndex + 1 < cells.Count && rightBorder)
{
visibleBorders.Add(cellIndex + 1);
}
cellIndex += 1;
}
return(visibleBorders.Count);
}
void CollectRows(WordDocHolder docHolder, int maxRowsToProcess, string extension)
{
var docPart = docHolder.WordDocument.MainDocumentPart;
var tables = docPart.Document.Descendants <Table>().ToList();
int tableIndex = 0;
foreach (OpenXmlPart h in docPart.HeaderParts)
{
foreach (var t in h.RootElement.Descendants <Table>())
{
ProcessWordTableAndUpdateTitle(docHolder, t, maxRowsToProcess, tableIndex);
tableIndex++;
}
}
if (extension != ".htm" && extension != ".html") // это просто костыль. Нужно как-то встроить это в архитектуру.
{
tables = ExtractSubtables(tables);
}
TablesCount = tables.Count();
foreach (var t in tables)
{
ProcessWordTableAndUpdateTitle(docHolder, t, maxRowsToProcess, tableIndex);
tableIndex++;
}
TableRows = DropDayOfWeekRows(TableRows);
}
void ProcessWordTableAndUpdateTitle(WordDocHolder docHolder, Table table, int maxRowsToProcess, int tableIndex)
{
if (CanAppendRowsFromTable(table, tableIndex))
{
ProcessWordTable(docHolder, table, maxRowsToProcess);
}
int debugSaveRowCount = TableRows.Count;
if (TableRows.Count > debugSaveRowCount)
{
string tableText = table.InnerText.Length > 30 ? table.InnerText.Substring(0, 30) : table.InnerText;
Logger.Debug(String.Format("add {0} rows (TableRows.Count={1} ) from table {2} Table.innertText[0:30]='{3}'",
TableRows.Count - debugSaveRowCount,
TableRows.Count,
tableIndex,
tableText));
}
if (Title.Length == 0 && table.InnerText.Length > 30 && table.InnerText.ToLower().IndexOf("декабря") != -1)
{
var rows = new List <String>();
foreach (var r in table.Descendants <TableRow>())
{
rows.Add(r.InnerText);
}
Title = String.Join("\n", rows);
}
}
TableWidthInfo InitializeTableWidthInfo(WordDocHolder docHolder, Table table)
{
TableWidthInfo widthInfo = new TableWidthInfo();
TableProperties tProp = table.GetFirstChild <TableProperties>();
if (tProp != null)
{
if (tProp.TableWidth != null)
{
widthInfo.TableWidthInPixels = TableWidthInfo.TryReadWidth(
tProp.TableWidth.Width,
tProp.TableWidth.Type,
docHolder.DocumentPageSizeInPixels);
}
if (tProp.TableIndentation != null)
{
widthInfo.TableIndentionInPixels = TableWidthInfo.TryReadWidth(
tProp.TableIndentation.Width,
tProp.TableIndentation.Type,
docHolder.DocumentPageSizeInPixels);
}
widthInfo.TableIndentionInPixels += docHolder.DocumentPageLeftMaginInPixels;
}
else
{
widthInfo.TableWidthInPixels = docHolder.DocumentPageSizeInPixels;
}
TableGrid tGrid = table.GetFirstChild <TableGrid>();
if (tGrid != null)
{
widthInfo.ColumnWidths = new List <int>();
foreach (var col in tGrid.Elements <GridColumn>())
{
widthInfo.ColumnWidths.Add(
TableWidthInfo.TryReadWidth(
col.Width,
TableWidthUnitValues.Dxa,
widthInfo.TableWidthInPixels));
}
}
return(widthInfo);
}
private void ProcessDoc(string fileName, string extension, int maxRowsToProcess)
{
using (var doc = new WordDocHolder(WordprocessingDocument.Open(fileName, false)))
{
CurrentScheme = _allSchemes.Find(x => x.CanProcess(doc.WordDocument));
if (CurrentScheme != default)
{
// CollectRows from distinct Tables
Title = doc.FindTitleAboveTheTable();
CurrentScheme.Document = doc.WordDocument.MainDocumentPart.Document;
TablesCount = 1;
}
else
{
Title = doc.FindTitleAboveTheTable();
CollectRows(doc, maxRowsToProcess, extension);
UnmergedColumnsCount = GetUnmergedColumnsCountByFirstRow();
FindRowsThatCannotBeMerged();
InitializeVerticallyMerge();
}
};
}
void ProcessWordTable(WordDocHolder docHolder, Table table, int maxRowsToProcess)
{
var rows = table.Descendants <TableRow>().ToList();
TableWidthInfo widthInfo = InitializeTableWidthInfo(docHolder, table);
int saveRowsCount = TableRows.Count;
int maxCellsCount = 0;
TableBorders tblBorders = GetTableBorders(table);
for (int r = 0; r < rows.Count(); ++r)
{
OpenXmlTableRow newRow = new OpenXmlTableRow();
int sumspan = 0;
var tableRow = rows[r];
int rowGridBefore = GetRowGridBefore(tableRow);
bool isEmpty = true;
var row = tableRow.Elements <TableCell>().ToArray();
for (var i = 0; i < row.Length; ++i)
{
var c = new OpenXmlWordCell(docHolder, row, i, widthInfo, TableRows.Count, sumspan, tblBorders);
if (newRow.RowCells.Count == 0)
{
c.MergedColsCount += rowGridBefore;
}
if (newRow.RowCells.Count > 0 && !newRow.RowCells.Last().HasRightBorder)
{
newRow.RowCells.Last().Text += c.Text;
newRow.RowCells.Last().CellWidth += c.CellWidth;
newRow.RowCells.Last().MergedColsCount += c.MergedColsCount;
newRow.RowCells.Last().HasRightBorder = c.HasRightBorder;
sumspan += c.MergedColsCount;
}
else
{
newRow.RowCells.Add(c);
sumspan += c.MergedColsCount;
}
isEmpty = isEmpty && c.IsEmpty;
}
if (isEmpty)
{
continue;
}
maxCellsCount = Math.Max(newRow.RowCells.Count, maxCellsCount);
if (r == 0 && TableRows.Count > 0 &&
BigramsHolder.CheckMergeRow(
TableRows.Last().RowCells.ConvertAll(x => x.Text),
newRow.RowCells.ConvertAll(x => x.Text)))
{
MergeRow(TableRows.Last().RowCells, newRow.RowCells);
}
else
{
TableRows.Add(newRow);
}
if ((maxRowsToProcess != -1) && (TableRows.Count >= maxRowsToProcess))
{
break;
}
}
if ((TableRows.Count > 0) && !TableHeaderRecognizer.IsNamePositionAndIncomeTable(GetDataCells(0)))
{
if (maxCellsCount <= 4 || CheckNameColumnIsEmpty(saveRowsCount))
{
//remove this suspicious table
TableRows.RemoveRange(saveRowsCount, TableRows.Count - saveRowsCount);
}
}
}
private void InitTextProperties(WordDocHolder docHolder, OpenXmlElement inputCell)
{
Text = "";
foreach (var p in inputCell.Elements <Paragraph>())
{
foreach (var textOrBreak in p.Descendants())
{
if (textOrBreak.LocalName == "r" && textOrBreak is Run)
{
Run r = textOrBreak as Run;
RunProperties rProps = r.RunProperties;
if (rProps != null)
{
if (rProps.FontSize != null)
{
int runFontSize = Int32.Parse(rProps.FontSize.Val);
if (runFontSize <= 28)
{
FontSize = runFontSize; // if font is too large, it is is an ocr error, ignore it
}
}
if (rProps.RunFonts != null)
{
FontName = rProps.RunFonts.ComplexScript;
}
}
}
else if (textOrBreak.LocalName == "t")
{
Text += textOrBreak.InnerText;
}
else if (textOrBreak.LocalName == "cr")
{
Text += "\n";
}
else if (textOrBreak.LocalName == "br")
/* do not use lastRenderedPageBreak, see MinRes2011 for wrong lastRenderedPageBreak in Семенов
||
|| (textOrBreak.Name == w + "lastRenderedPageBreak") */
{
Text += "\n";
}
else if (textOrBreak.LocalName == "numPr")
{
Text += "- ";
}
}
Text += "\n";
ParagraphProperties pPr = p.ParagraphProperties;
if (pPr != null)
{
for (int l = 0; l < AfterLinesCount(pPr.SpacingBetweenLines); ++l)
{
Text += "\n";
}
}
}
IsEmpty = Text.IsNullOrWhiteSpace();
if (string.IsNullOrEmpty(FontName))
{
FontName = docHolder.DefaultFontName;
}
if (FontSize == 0)
{
FontSize = docHolder.DefaultFontSize;
}
}
public OpenXmlWordCell(WordDocHolder docHolder, TableCell[] row, int cellIndexInRow, TableWidthInfo tableWidth,
int rowIndexInTable, int unmergedColumnIndex, TableBorders tblBorders)
{
TableCell inputCell = row[cellIndexInRow];
InitTextProperties(docHolder, inputCell);
if (inputCell?.TableCellProperties?.TableCellBorders != null)
{
var borders = inputCell.TableCellProperties.TableCellBorders;
HasBottomBorder = WordDocHolder.BorderIsVisible(borders.BottomBorder);
HasTopBorder = WordDocHolder.BorderIsVisible(borders.TopBorder);
HasRightBorder = WordDocHolder.BorderIsVisible(borders.RightBorder);
if (!HasRightBorder && cellIndexInRow + 1 < row.Length && row[cellIndexInRow + 1].TableCellProperties?.TableCellBorders != null)
{
HasRightBorder = WordDocHolder.BorderIsVisible(row[cellIndexInRow + 1].TableCellProperties?.TableCellBorders?.LeftBorder);
}
if (!HasRightBorder && tblBorders?.InsideVerticalBorder != null && (uint)tblBorders.InsideVerticalBorder.Size > 0)
{
HasRightBorder = true;
}
}
var vmerge = inputCell.TableCellProperties.GetFirstChild <VerticalMerge>();
VerticallyMerged = null;
if (vmerge != null)
{
if ((vmerge.Val == null) || (vmerge.Val == MergedCellValues.Continue))
{
// null -> MergedCellValues.Continue
VerticallyMerged = MergedCellValues.Continue;
}
else if (vmerge.Val == MergedCellValues.Restart)
{
VerticallyMerged = MergedCellValues.Restart;
}
}
if (tblBorders?.InsideHorizontalBorder != null && (uint)tblBorders.InsideHorizontalBorder.Size > 0)
{
TableHasInsideHorizontalBorders = true;
}
var gridSpan = inputCell.TableCellProperties.GetFirstChild <GridSpan>();
IsMerged = gridSpan != null && gridSpan.Val > 1;
FirstMergedRow = -1; // init afterwards
MergedRowsCount = -1; // init afterwards
MergedColsCount = (gridSpan == null) ? 1 : (int)gridSpan.Val;
Row = rowIndexInTable;
Col = unmergedColumnIndex;
if (inputCell.TableCellProperties != null &&
inputCell.TableCellProperties.TableCellWidth != null &&
inputCell.TableCellProperties.TableCellWidth.Type != null &&
inputCell.TableCellProperties.TableCellWidth.Type != TableWidthUnitValues.Auto
)
{
CellWidth = TableWidthInfo.TryReadWidth(
inputCell.TableCellProperties.TableCellWidth.Width,
inputCell.TableCellProperties.TableCellWidth.Type,
tableWidth.TableWidthInPixels);
}
else
{
if (Col < tableWidth.ColumnWidths.Count)
{
CellWidth = tableWidth.ColumnWidths[Col];
}
}
AdditTableIndention = tableWidth.TableIndentionInPixels;
}
