/// <summary>
/// Performs a parsing pass over a tbody to read information about column width and rowspan attributes. Information read about width
/// attributes is stored in the reference ArrayList parameter columnStarts, which contains a list of all starting
/// positions of all columns in the table, ordered from left to right. Row spans are taken into consideration when
/// computing column starts
/// </summary>
/// <param name="htmlTbodyElement">
/// XmlElement representing Html tbody whose structure is to be analyzed
/// </param>
/// <param name="columnStarts">
/// ArrayList of type double which contains the function output. If analysis fails, this parameter is set to null
/// </param>
/// <param name="tableWidth">
/// Current width of the table. This is used to determine if a new column when added to the end of table should
/// come after the last column in the table or is actually splitting the last column in two. If it is only splitting
/// the last column it should inherit row span for that column
/// </param>
/// <returns>
/// Calculated width of a tbody.
/// In case of non-analizable column width structure return 0;
/// </returns>
private static double AnalyzeTbodyStructure(IXmlNode htmlTbodyElement, List<double> columnStarts, List<int> activeRowSpans, double tableWidth)
{
// Parameter validation
Debug.Assert(htmlTbodyElement.LocalName.ToLower() == "tbody");
Debug.Assert(columnStarts != null);
double tbodyWidth = 0;
bool columnWidthsAvailable = true;
if (!htmlTbodyElement.HasChildNodes())
{
return tbodyWidth;
}
// Set active row spans to 0 - thus ignoring row spans crossing tbody boundaries
ClearActiveRowSpans(activeRowSpans);
var htmlChildNode = htmlTbodyElement.FirstChild;
// Analyze tr elements
while (htmlChildNode != null && columnWidthsAvailable)
{
switch (htmlChildNode.LocalName.ToLower())
{
case "tr":
double trWidth = AnalyzeTRStructure(htmlChildNode, columnStarts, activeRowSpans, tbodyWidth);
if (trWidth > tbodyWidth)
{
tbodyWidth = trWidth;
}
break;
case "td":
columnWidthsAvailable = false; // interrupt the analisys
break;
default:
break;
}
htmlChildNode = htmlChildNode.NextSibling;
}
// Set active row spans to 0 - thus ignoring row spans crossing tbody boundaries
ClearActiveRowSpans(activeRowSpans);
return columnWidthsAvailable ? tbodyWidth : 0;
}
/// <summary>
/// Performs a parsing pass over a tr element to read information about column width and rowspan attributes.
/// </summary>
/// <param name="htmlTRElement">
/// XmlElement representing Html tr element whose structure is to be analyzed
/// </param>
/// <param name="columnStarts">
/// ArrayList of type double which contains the function output. If analysis is successful, this ArrayList contains
/// all the points which are the starting position of any column in the tr, ordered from left to right. If analysis fails,
/// the ArrayList is set to null
/// </param>
/// <param name="activeRowSpans">
/// ArrayList representing all columns currently spanned by an earlier row span attribute. These columns should
/// not be used for data in this row. The ArrayList actually contains notation for all columns in the table, if the
/// active row span is set to 0 that column is not presently spanned but if it is > 0 the column is presently spanned
/// </param>
/// <param name="tableWidth">
/// Double value representing the current width of the table.
/// Return 0 if analisys was insuccessful.
/// </param>
private static double AnalyzeTRStructure(IXmlNode htmlTRElement, List<double> columnStarts, List<int> activeRowSpans, double tableWidth)
{
double columnWidth;
// Parameter validation
Debug.Assert(htmlTRElement.LocalName.ToLower() == "tr");
Debug.Assert(columnStarts != null);
Debug.Assert(activeRowSpans != null);
Debug.Assert(columnStarts.Count == activeRowSpans.Count);
if (!htmlTRElement.HasChildNodes())
{
return 0;
}
bool columnWidthsAvailable = true;
double columnStart = 0; // starting position of current column
var htmlChildNode = htmlTRElement.FirstChild;
int columnIndex = 0;
double trWidth = 0;
// Skip spanned columns to get to real column start
if (columnIndex < activeRowSpans.Count)
{
Debug.Assert((double)columnStarts[columnIndex] >= columnStart);
if ((double)columnStarts[columnIndex] == columnStart)
{
// The new column may be in a spanned area
while (columnIndex < activeRowSpans.Count && (int)activeRowSpans[columnIndex] > 0)
{
activeRowSpans[columnIndex] = (int)activeRowSpans[columnIndex] - 1;
Debug.Assert((int)activeRowSpans[columnIndex] >= 0);
columnIndex++;
columnStart = (double)columnStarts[columnIndex];
}
}
}
while (htmlChildNode != null && columnWidthsAvailable)
{
Debug.Assert(columnStarts.Count == activeRowSpans.Count);
VerifyColumnStartsAscendingOrder(columnStarts);
switch (htmlChildNode.LocalName.ToLower())
{
case "td":
Debug.Assert(columnIndex <= columnStarts.Count);
if (columnIndex < columnStarts.Count)
{
Debug.Assert(columnStart <= (double)columnStarts[columnIndex]);
if (columnStart < (double)columnStarts[columnIndex])
{
columnStarts.Insert(columnIndex, columnStart);
// There can be no row spans now - the column data will appear here
// Row spans may appear only during the column analysis
activeRowSpans.Insert(columnIndex, 0);
}
}
else
{
// Column start is greater than all previous starts. Row span must still be 0 because
// we are either adding after another column of the same row, in which case it should not inherit
// the previous column's span. Otherwise we are adding after the last column of some previous
// row, and assuming the table widths line up, we should not be spanned by it. If there is
// an incorrect tbale structure where a columns starts in the middle of a row span, we do not
// guarantee correct output
columnStarts.Add(columnStart);
activeRowSpans.Add(0);
}
columnWidth = GetColumnWidth(htmlChildNode);
if (columnWidth != -1)
{
int nextColumnIndex;
int rowSpan = GetRowSpan(htmlChildNode);
nextColumnIndex = GetNextColumnIndex(columnIndex, columnWidth, columnStarts, activeRowSpans);
if (nextColumnIndex != -1)
{
// Entire column width can be processed without hitting conflicting row span. This means that
// column widths line up and we can process them
Debug.Assert(nextColumnIndex <= columnStarts.Count);
// Apply row span to affected columns
for (int spannedColumnIndex = columnIndex; spannedColumnIndex < nextColumnIndex; spannedColumnIndex++)
{
activeRowSpans[spannedColumnIndex] = rowSpan - 1;
Debug.Assert((int)activeRowSpans[spannedColumnIndex] >= 0);
}
columnIndex = nextColumnIndex;
// Calculate columnsStart for the next cell
columnStart = columnStart + columnWidth;
if (columnIndex < activeRowSpans.Count)
{
Debug.Assert((double)columnStarts[columnIndex] >= columnStart);
if ((double)columnStarts[columnIndex] == columnStart)
{
// The new column may be in a spanned area
while (columnIndex < activeRowSpans.Count && (int)activeRowSpans[columnIndex] > 0)
{
activeRowSpans[columnIndex] = (int)activeRowSpans[columnIndex] - 1;
Debug.Assert((int)activeRowSpans[columnIndex] >= 0);
columnIndex++;
columnStart = (double)columnStarts[columnIndex];
}
}
// else: the new column does not start at the same time as a pre existing column
// so we don't have to check it for active row spans, it starts in the middle
// of another column which has been checked already by the GetNextColumnIndex function
}
}
else
{
// Full column width cannot be processed without a pre existing row span.
// We cannot analyze widths
columnWidthsAvailable = false;
}
}
else
{
// Incorrect column width, stop processing
columnWidthsAvailable = false;
}
break;
default:
break;
}
htmlChildNode = htmlChildNode.NextSibling;
}
// The width of the tr element is the position at which it's last td element ends, which is calculated in
// the columnStart value after each td element is processed
if (columnWidthsAvailable)
{
trWidth = columnStart;
}
else
{
trWidth = 0;
}
return trWidth;
}
/// <summary>
/// Performs a parsing pass over a table to read information about column width and rowspan attributes. This information
/// is used to determine the starting point of each column.
/// </summary>
/// <param name="htmlTableElement">
/// XmlElement representing Html table whose structure is to be analyzed
/// </param>
/// <returns>
/// ArrayList of type double which contains the function output. If analysis is successful, this ArrayList contains
/// all the points which are the starting position of any column in the table, ordered from left to right.
/// In case if analisys was impossible we return null.
/// </returns>
private static List<double> AnalyzeTableStructure(IXmlNode htmlTableElement)
{
// Parameter validation
Debug.Assert(htmlTableElement.LocalName.ToLower() == "table");
if (!htmlTableElement.HasChildNodes())
{
return null;
}
bool columnWidthsAvailable = true;
var columnStarts = new List<double>();
var activeRowSpans = new List<int>();
Debug.Assert(columnStarts.Count == activeRowSpans.Count);
var htmlChildNode = htmlTableElement.FirstChild;
double tableWidth = 0; // Keep track of table width which is the width of its widest row
// Analyze tbody and tr elements
while (htmlChildNode != null && columnWidthsAvailable)
{
Debug.Assert(columnStarts.Count == activeRowSpans.Count);
switch (htmlChildNode.LocalName.ToLower())
{
case "tbody":
// Tbody element, we should analyze its children for trows
double tbodyWidth = AnalyzeTbodyStructure(htmlChildNode, columnStarts, activeRowSpans, tableWidth);
if (tbodyWidth > tableWidth)
{
// Table width must be increased to supported newly added wide row
tableWidth = tbodyWidth;
}
else if (tbodyWidth == 0)
{
// Tbody analysis may return 0, probably due to unprocessable format.
// We should also fail.
columnWidthsAvailable = false; // interrupt the analisys
}
break;
case "tr":
// Table row. Analyze column structure within row directly
double trWidth = AnalyzeTRStructure(htmlChildNode, columnStarts, activeRowSpans, tableWidth);
if (trWidth > tableWidth)
{
tableWidth = trWidth;
}
else if (trWidth == 0)
{
columnWidthsAvailable = false; // interrupt the analisys
}
break;
case "td":
// Incorrect formatting, too deep to analyze at this level. Return null.
// TODO: implement analysis at this level, possibly by creating a new tr
columnWidthsAvailable = false; // interrupt the analisys
break;
default:
// Element should not occur directly in table. Ignore it.
break;
}
htmlChildNode = htmlChildNode.NextSibling;
}
if (columnWidthsAvailable)
{
// Add an item for whole table width
columnStarts.Add(tableWidth);
VerifyColumnStartsAscendingOrder(columnStarts);
}
else
{
columnStarts = null;
}
return columnStarts;
}
