Files
speckle-sharp-connectors/Converters/CSi/Speckle.Converters.CSiShared/Extensions/DatabaseTableExtensions.cs
T
Björn Steinhagen 0e65b6aa57 feat(etabs): geometric properties through new cDatabaseTable extractor (#584)
* feature: database table extractor

* feature: frame length extraction

* feat: shell area extraction

* feat: frame length extraction

* feat: shell area extraction

* docs: cleaned up object extractors

* refactor: extensible database table extraction

* flexible indexing (can be defined by user, defaults to typical use-case of "UniqueName")
* better variable naming and more descriptive (better than csi's api)
* some fail safes e.g. when forgetting to put indexingColumn in array of columns to fetch

* refactor: using DatabaseTableExtractor for material name lookup

* lookup for material name of a shell can use the database table extractor and simplifies EtabsShellPropertiesExtractor a lot

* fix: openings shouldn't look for material assignments

Frustrating way Etabs handles openings - they shouldn't be a SHELL. Anyways. Rant over

* refactor: database table extractor handles dictionary lookups
2025-02-12 18:43:44 +00:00

130 lines
5.3 KiB
C#

namespace Speckle.Converters.CSiShared.Extensions;
/// <summary>
/// Csi Api returns a one-dimensional array of the table data. Any cDatabaseTable queries will require some processing.
/// The TableData extension processes queries for GetTableForDisplayArray.
/// </summary>
/// <remarks>
/// TableData implemented as a record. Reasons for this include:
/// <list type="bullet">
/// <item><description>Keeping data immutable (preventing accidental modifications).</description></item>
/// <item><description>Better choice for large data sets (heap allocation).</description></item>
/// </list>
/// Notes:
/// <list type="bullet">
/// <item><description>A cDatabaseTable query returns ALL objects of a type. This is an expensive operation. However, the typical use-case involves sending the entire Etabs/Sap model.</description></item>
/// <item><description>High initial memory usage when creating dictionaries for all rows of data.</description></item>
/// <item><description>Benefits of the dictionary evident during send operations when most/all objects are sent (and thus queried).</description></item>
/// <item><description>Single upfront dictionary creation preferred over repeated on-demand creation</description></item>
/// <item><description>Yes, Csi returns all data as strings. Even int, double etc.</description></item>
/// </list>
/// </remarks>
public record TableData
{
private readonly string[] _columnNames; // "fieldKeys" in api docs
private readonly string[] _rawTableData; // indicating raw, one-dimensional array of table data (before processing)
private readonly int _rowCount; // Number of rows
private IReadOnlyDictionary<string, IReadOnlyDictionary<string, string>>? _processedRows; // Cached data structure
private readonly string _indexColumn; // column used to index/identify rows (typically, "UniqueName")
public TableData(string[] columnNames, string[] rawTableData, int rowCount, string indexColumn)
{
_columnNames = columnNames;
_rawTableData = rawTableData;
_rowCount = rowCount;
_indexColumn = indexColumn;
}
/// <summary>
/// Gets table data as a dictionary mapping indexColumn (typically "UniqueName" to _processedRows).
/// Each row is itself a dictionary mapping column names to their values.
/// Computed once on first access and cached.
/// </summary>
/// <remarks>
/// Motivation:
/// <list type="bullet">
/// <item><description>One-dimensional array => structured dictionary format</description></item>
/// <item><description>Each row keyed by its "UniqueName" value</description></item>
/// <item><description>Each row value is itself a dictionary of field keys to values</description></item>
/// </list>
/// </remarks>
public IReadOnlyDictionary<string, IReadOnlyDictionary<string, string>> Rows
{
get
{
if (_processedRows != null) // Lazy loading - only build dictionary when first accessed
{
return _processedRows;
}
var columnsPerRow = _columnNames.Length;
var indexColumnIndex = Array.IndexOf(_columnNames, _indexColumn);
if (indexColumnIndex == -1)
{
throw new InvalidOperationException(
$"Row data structured according specified '{_indexColumn}' field. This was not found in the database."
);
}
// Pre-size dictionary with known capacity
var rows = new Dictionary<string, IReadOnlyDictionary<string, string>>(_rowCount);
// Create a field index lookup to avoid repeated Array.IndexOf calls
var fieldIndexLookup = new Dictionary<string, int>(columnsPerRow);
for (int i = 0; i < _columnNames.Length; i++)
{
fieldIndexLookup[_columnNames[i]] = i;
}
// Process each row
for (int rowStart = 0; rowStart < _rawTableData.Length; rowStart += columnsPerRow)
{
var keyValue = _rawTableData[rowStart + indexColumnIndex];
// Pre-size the row dictionary
var row = new Dictionary<string, string>(columnsPerRow, StringComparer.Ordinal);
// Use index lookup instead of repeated string comparisons
foreach (var kvp in fieldIndexLookup)
{
row[kvp.Key] = _rawTableData[rowStart + kvp.Value];
}
rows[keyValue] = row;
}
_processedRows = rows;
return _processedRows;
}
}
/// <summary>
/// Retrieves a string value from a specific row and column from the table data.
/// </summary>
/// <param name="rowKey">The unique identifier for the row, matching the value in the index column (e.g., "UniqueName")</param>
/// <param name="columnName">The name of the column containing the desired value</param>
/// <returns>The string value found at the specified row and column intersection</returns>
/// <exception cref="InvalidOperationException">Thrown when either the row or column is not found in the table</exception>
public string GetRowValue(string rowKey, string columnName)
{
if (TryGetValue(rowKey, columnName, out var value))
{
return value;
}
throw new InvalidOperationException($"Failed to get value for row '{rowKey}', column '{columnName}'");
}
private bool TryGetValue(string rowKey, string columnName, out string value)
{
if (Rows.TryGetValue(rowKey, out var row) && row.TryGetValue(columnName, out value!))
{
return true;
}
value = string.Empty;
return false;
}
}