Implement phase 4 critical table imports

This commit is contained in:
2026-03-14 03:27:14 +01:00
parent a391a1421a
commit b2f61c3d73
17 changed files with 1280 additions and 474 deletions

View File

@@ -0,0 +1,276 @@
namespace RolemasterDb.ImportTool.Parsing;
public sealed class VariantColumnCriticalTableParser
{
private static readonly ColumnDefinition[] ExpectedColumns =
[
new("NORMAL", "Normal"),
new("MAGIC", "Magic"),
new("MITHRIL", "Mithril"),
new("HOLY_ARMS", "Holy Arms"),
new("SLAYING", "Slaying")
];
public CriticalTableParseResult Parse(CriticalImportManifestEntry entry, string xmlContent)
{
var fragments = CriticalTableParserSupport.LoadFragments(xmlContent);
var headerFragments = FindHeaderFragments(fragments);
var validationErrors = new List<string>();
var validationWarnings = new List<string>();
var columnAnchors = headerFragments
.OrderBy(item => item.Left)
.Select(item =>
{
var definition = ResolveColumnDefinition(item.Text);
return (definition.Key, definition.Label, item.CenterX);
})
.ToList();
var bodyStartTop = headerFragments.Max(item => item.Top) + CriticalTableParserSupport.HeaderToBodyMinimumGap;
var keyTop = fragments
.Where(item =>
string.Equals(item.Text, "Key:", StringComparison.OrdinalIgnoreCase) ||
item.Text.Contains("must parry", StringComparison.OrdinalIgnoreCase) ||
item.Text.Contains("attacker gets", StringComparison.OrdinalIgnoreCase))
.Select(item => (int?)item.Top)
.Min() ?? int.MaxValue;
var affixLegendSymbols = CriticalTableParserSupport.DetectAffixLegendSymbols(fragments, keyTop);
var leftCutoff = headerFragments.Min(item => item.Left) - 10;
var rowLabelFragments = CriticalTableParserSupport.FindRowLabelFragments(
fragments,
leftCutoff,
bodyStartTop,
keyTop);
var rowAnchors = rowLabelFragments
.OrderBy(item => item.Top)
.Select((item, index) => new RowAnchor(CriticalTableParserSupport.NormalizeRollBandLabel(item.Text), item.Top, index + 1))
.ToList();
if (rowAnchors.Count == 0)
{
validationErrors.Add("No roll-band labels were found in the XML artifact.");
}
var columnCenters = columnAnchors
.Select(item => (item.Key, item.CenterX))
.ToList();
var bodyFragments = fragments
.Where(item =>
item.Top >= bodyStartTop &&
item.Top < keyTop - CriticalTableParserSupport.TopGroupingTolerance &&
!CriticalTableParserSupport.IsFooterPageNumberFragment(item, keyTop) &&
!CriticalTableParserSupport.IsPotentialRowLabelFragment(item, leftCutoff) &&
!rowAnchors.Any(anchor => anchor.Top == item.Top && string.Equals(anchor.Label, CriticalTableParserSupport.NormalizeRollBandLabel(item.Text), StringComparison.OrdinalIgnoreCase)) &&
!headerFragments.Contains(item))
.ToList();
bodyFragments = CriticalTableParserSupport.SplitBoundaryCrossingAffixFragments(bodyFragments, columnCenters, affixLegendSymbols);
var bodyLines = CriticalTableParserSupport.BuildBodyLines(bodyFragments, columnCenters, affixLegendSymbols);
var parsedRollBands = rowAnchors
.Select(anchor => CriticalTableParserSupport.CreateRollBand(anchor.Label, anchor.SortOrder))
.ToList();
var cellEntries = new List<CellEntry>();
for (var rowIndex = 0; rowIndex < rowAnchors.Count; rowIndex++)
{
var rowStart = rowIndex == 0
? bodyStartTop
: ResolveRowBoundaryTop(rowAnchors[rowIndex - 1], rowAnchors[rowIndex], bodyLines);
var rowEnd = rowIndex == rowAnchors.Count - 1
? keyTop - 1
: ResolveRowBoundaryTop(rowAnchors[rowIndex], rowAnchors[rowIndex + 1], bodyLines);
var rowFragments = bodyFragments
.Where(item => item.Top >= rowStart && item.Top < rowEnd)
.ToList();
foreach (var columnAnchor in columnAnchors)
{
var cellFragments = rowFragments
.Where(item => CriticalTableParserSupport.ResolveColumn(item.CenterX, columnCenters) == columnAnchor.Key)
.OrderBy(item => item.Top)
.ThenBy(item => item.Left)
.ToList();
if (cellFragments.Count == 0)
{
validationErrors.Add($"Missing content for roll band '{rowAnchors[rowIndex].Label}', column '{columnAnchor.Key}'.");
continue;
}
cellEntries.Add(new CellEntry(
rowAnchors[rowIndex].Label,
rowIndex,
columnAnchor.Key,
CriticalTableParserSupport.BuildLines(cellFragments).ToList()));
}
}
RepairLeadingAffixLeakage(cellEntries, affixLegendSymbols);
var parsedCells = new List<ParsedCriticalCellArtifact>();
var parsedResults = new List<ParsedCriticalResult>();
foreach (var cellEntry in cellEntries.OrderBy(item => item.RowIndex).ThenBy(item => item.ColumnKey, StringComparer.Ordinal))
{
var segmentCount = CriticalTableParserSupport.CountLineTypeSegments(cellEntry.Lines, affixLegendSymbols);
if (segmentCount > 2)
{
validationErrors.Add($"Cell '{cellEntry.RollBandLabel}/{cellEntry.ColumnKey}' interleaves prose and affix lines.");
}
var rawAffixLines = cellEntry.Lines.Where(line => CriticalTableParserSupport.IsAffixLikeLine(line, affixLegendSymbols)).ToList();
var descriptionLines = cellEntry.Lines.Where(line => !CriticalTableParserSupport.IsAffixLikeLine(line, affixLegendSymbols)).ToList();
var rawCellText = string.Join(Environment.NewLine, cellEntry.Lines);
var descriptionText = CriticalTableParserSupport.CollapseWhitespace(string.Join(' ', descriptionLines));
var rawAffixText = rawAffixLines.Count == 0 ? null : string.Join(Environment.NewLine, rawAffixLines);
parsedCells.Add(new ParsedCriticalCellArtifact(
null,
cellEntry.RollBandLabel,
cellEntry.ColumnKey,
cellEntry.Lines,
rawCellText,
descriptionText,
rawAffixText));
parsedResults.Add(new ParsedCriticalResult(
null,
cellEntry.ColumnKey,
cellEntry.RollBandLabel,
rawCellText,
descriptionText,
rawAffixText));
}
if (columnAnchors.Count != ExpectedColumns.Length)
{
validationErrors.Add($"Expected {ExpectedColumns.Length} variant columns but found {columnAnchors.Count}.");
}
if (parsedCells.Count != rowAnchors.Count * columnAnchors.Count)
{
validationErrors.Add($"Expected {rowAnchors.Count * columnAnchors.Count} parsed cells but produced {parsedCells.Count}.");
}
var validationReport = new ImportValidationReport(
validationErrors.Count == 0,
validationErrors,
validationWarnings,
rowAnchors.Count,
parsedCells.Count);
var table = new ParsedCriticalTable(
entry.Slug,
entry.DisplayName,
entry.Family,
Path.GetFileName(entry.PdfPath),
"Imported from PDF XML extraction.",
[],
ExpectedColumns.Select((item, index) => new ParsedCriticalColumn(item.Key, item.Label, "variant", index + 1)).ToList(),
parsedRollBands,
parsedResults);
return new CriticalTableParseResult(table, fragments, parsedCells, validationReport);
}
private static List<XmlTextFragment> FindHeaderFragments(IReadOnlyList<XmlTextFragment> fragments)
{
var expectedLabels = ExpectedColumns
.Select(item => item.Label.ToLowerInvariant())
.ToList();
var headerCandidates = fragments
.Where(item => expectedLabels.Contains(item.Text.Trim().ToLowerInvariant(), StringComparer.Ordinal))
.OrderBy(item => item.Top)
.ThenBy(item => item.Left)
.ToList();
foreach (var group in CriticalTableParserSupport.GroupByTop(headerCandidates))
{
var ordered = group.OrderBy(item => item.Left).ToList();
var labels = ordered.Select(item => item.Text.Trim().ToLowerInvariant()).ToList();
if (labels.SequenceEqual(expectedLabels))
{
return ordered;
}
}
throw new InvalidOperationException("Could not find the variant-column header row in the XML artifact.");
}
private static ColumnDefinition ResolveColumnDefinition(string value) =>
ExpectedColumns.SingleOrDefault(item => string.Equals(item.Label, value.Trim(), StringComparison.OrdinalIgnoreCase))
?? throw new InvalidOperationException($"Unsupported variant column label '{value}'.");
private static void RepairLeadingAffixLeakage(List<CellEntry> cellEntries, ISet<string> affixLegendSymbols)
{
var maxRowIndex = cellEntries.Count == 0 ? -1 : cellEntries.Max(item => item.RowIndex);
var columnKeys = cellEntries.Select(item => item.ColumnKey).Distinct(StringComparer.OrdinalIgnoreCase).ToList();
for (var rowIndex = 0; rowIndex < maxRowIndex; rowIndex++)
{
foreach (var columnKey in columnKeys)
{
var current = cellEntries.SingleOrDefault(item => item.RowIndex == rowIndex && item.ColumnKey == columnKey);
var next = cellEntries.SingleOrDefault(item => item.RowIndex == rowIndex + 1 && item.ColumnKey == columnKey);
if (current is null || next is null)
{
continue;
}
var leadingAffixCount = 0;
while (leadingAffixCount < next.Lines.Count && CriticalTableParserSupport.IsAffixLikeLine(next.Lines[leadingAffixCount], affixLegendSymbols))
{
leadingAffixCount++;
}
if (leadingAffixCount == 0 || leadingAffixCount == next.Lines.Count)
{
continue;
}
current.Lines.AddRange(next.Lines.Take(leadingAffixCount));
next.Lines.RemoveRange(0, leadingAffixCount);
}
}
}
private static int ResolveRowBoundaryTop(
RowAnchor current,
RowAnchor next,
IReadOnlyList<(int Top, bool IsAffixLike)> bodyLines)
{
var linesBetweenLabels = bodyLines
.Where(item => item.Top >= current.Top && item.Top < next.Top)
.OrderBy(item => item.Top)
.ToList();
for (var index = linesBetweenLabels.Count - 2; index >= 0; index--)
{
if (linesBetweenLabels[index].IsAffixLike && !linesBetweenLabels[index + 1].IsAffixLike)
{
return (int)Math.Floor((linesBetweenLabels[index].Top + linesBetweenLabels[index + 1].Top) / 2.0) + 1;
}
}
return (int)Math.Floor((current.Top + next.Top) / 2.0) + 1;
}
private sealed record ColumnDefinition(string Key, string Label);
private sealed record RowAnchor(string Label, int Top, int SortOrder);
private sealed class CellEntry(string rollBandLabel, int rowIndex, string columnKey, List<string> lines)
{
public string RollBandLabel { get; } = rollBandLabel;
public int RowIndex { get; } = rowIndex;
public string ColumnKey { get; } = columnKey;
public List<string> Lines { get; } = lines;
}
}