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 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;
for (int r = 0; r < rows.Count(); ++r)
{
List <OpenXmlWordCell> newRow = new List <OpenXmlWordCell>();
int sumspan = 0;
var row = rows[r];
int rowGridBefore = GetRowGridBefore(row);
bool isEmpty = true;
foreach (var rowCell in row.Elements <TableCell>())
{
var c = new OpenXmlWordCell(docHolder, widthInfo, rowCell, TableRows.Count, sumspan);
if (newRow.Count == 0)
{
c.MergedColsCount += rowGridBefore;
}
newRow.Add(c);
sumspan += c.MergedColsCount;
isEmpty = isEmpty && c.IsEmpty;
}
if (isEmpty)
{
continue;
}
maxCellsCount = Math.Max(newRow.Count, maxCellsCount);
if (r == 0 && TableRows.Count > 0 &&
BigramsHolder.CheckMergeRow(
TableRows.Last().ConvertAll(x => x.Text),
newRow.ConvertAll(x => x.Text)))
{
MergeRow(TableRows.Last(), newRow);
}
else
{
TableRows.Add(newRow);
}
if ((maxRowsToProcess != -1) && (TableRows.Count >= maxRowsToProcess))
{
break;
}
}
if (maxCellsCount <= 4 || CheckNameColumnIsEmpty(TableRows, saveRowsCount))
{
//remove this suspicious table
TableRows.RemoveRange(saveRowsCount, TableRows.Count - saveRowsCount);
}
}
public OpenXmlWordCell(WordDocHolder docHolder, TableWidthInfo tableWidth, TableCell inputCell, int row, int column)
{
InitTextProperties(docHolder, inputCell);
var vmerge = inputCell.TableCellProperties.GetFirstChild <VerticalMerge>();
if (vmerge == null)
{
IsVerticallyMerged = false;
}
else
{
if (vmerge == null || vmerge.Val == null || vmerge.Val == MergedCellValues.Continue)
{
IsVerticallyMerged = true;
}
else
{
//vmerge.Val == MergedCellValues.Restart
IsVerticallyMerged = false;
}
}
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 = row;
Col = column;
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;
}
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();
InitializeVerticallyMerge();
}
};
}
void ProcessWordTableAndUpdateTitle(WordDocHolder docHolder, Table table, int maxRowsToProcess, int tableIndex)
{
int debugSaveRowCount = TableRows.Count;
if (table.Descendants <Table>().ToList().Count > 0)
{
Logger.Debug(String.Format("ignore table {0} with subtables", tableIndex));
}
else if (table.InnerText.Length > 0 && !table.InnerText.Any(x => Char.IsUpper(x)))
{
Logger.Debug(String.Format("ignore table {0} that has no uppercase char", tableIndex));
}
else if (table.InnerText.Length < 30)
{
Logger.Debug(String.Format("ignore table {0}, it is too short", tableIndex));
}
else
{
ProcessWordTable(docHolder, table, maxRowsToProcess);
}
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);
}
}
private void InitTextProperties(WordDocHolder docHolder, OpenXmlElement inputCell)
{
string s = "";
FontName = "";
FontSize = 0;
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")
{
s += textOrBreak.InnerText;
}
else if (textOrBreak.LocalName == "cr")
{
s += "\n";
}
else if (textOrBreak.LocalName == "br")
/* do not use lastRenderedPageBreak, see MinRes2011 for wrong lastRenderedPageBreak in Семенов
||
|| (textOrBreak.Name == w + "lastRenderedPageBreak") */
{
s += "\n";
}
else if (textOrBreak.LocalName == "numPr")
{
s += "- ";
}
}
s += "\n";
ParagraphProperties pPr = p.ParagraphProperties;
if (pPr != null)
{
for (int l = 0; l < AfterLinesCount(pPr.SpacingBetweenLines); ++l)
{
s += "\n";
}
}
}
Text = s;
IsEmpty = s.IsNullOrWhiteSpace();
if (string.IsNullOrEmpty(FontName))
{
FontName = docHolder.DefaultFontName;
}
if (FontSize == 0)
{
FontSize = docHolder.DefaultFontSize;
}
}
