Extends https://www.fluentassertions.com with specialized assertions for dealing with DataSet
, DataTable
, DataRow
and DataColumn
Warning
This set of extensions was extracted from Fluent Assertions 6 and is in maintenance mode. Pull Requests will still be accepted, but we will not actively work on improvements or bugs.
- Uses Semantic Versioning
- Uses Githubflow, so pull requests usually result in a new release
- The contribution guidelines are described here.
This package is shipped as a NuGet package and can be installed through the CLI using:
dotnet add package FluentAssertions.DataSets
- Visual Studio 2022 or JetBrains Rider 2021.3
- Build Tools 2022 (including the Universal Windows Platform build tools).
- .NET Framework 4.7 SDK
- .NET 7.0 SDK installed. Check the `global.json`` for the current minimum required version.
You can either build the solution using Visual Studio or Rider, or use the build.ps1
script to kick-off the Nuke build process. This will build the package and run all tests.
This is a special set of tests that use the Verify project to verify whether you've introduced any breaking changes in the public API of the library.
If you've verified the changes and decided they are valid, you can accept them using AcceptApiChanges.ps1
or AcceptApiChanges.sh
. Alternatively, you can use the Verify Support plug-in to compare the changes and accept them right from inside Rider. See also the Contribution Guidelines.
This package can be used to assert equivalence of System.Data
types such as DataSet
and DataTable
.
As with other reference types, you can assert a value of any of the core System.Data types to be null or not null:
DataSet result = ...;
result.Should().NotBeNull();
You can also assert that two DataSet
objects contain equivalent configuration and data, which, by default, will compare the rows contained by DataTable
objects by their index within the collection.
Like this:
var expected = GetExpectedDataSet();
var actual = GetActualDataSet();
actual.Should().BeEquivalentTo(expected);
The BeEquivalentTo
test can be applied to DataTable
, DataColumn
and DataRow
objects as well:
actual.Tables["First"].Should().BeEquivalentTo(expected.Tables["First"]);
Chaining additional assertions is supported as well.
dataTable.Should().HaveColumns("FirstName", "LastName")
.And.Should().HaveRowCount(3);
The following assertions are available on DataSet
objects:
.Should().HaveTableCount(n)
: Asserts that theDataSet
'sTables
collection has the expected number of members..Should().HaveTable(tableName)
,.Should().HaveTables(tableName, tableName, ...)
: Asserts that theDataSet
'sTables
collection contains at least tables with the specified names. Additional tables are ignored..Should().BeEquivalentTo(dataSet)
: Performs a deep equivalency comparison between the subject and the expectation.
The following assertions are available on DataTable
objects:
.Should().HaveColumn(columnName)
,.Should().HaveColumns(columnName, columnName, ...)
: Asserts that theDataTable
'sColumns
collection contains at least columns with the specified names. Additional columns are ignored..Should().HaveRowCount(n)
: Asserts that theDataTable
'sRows
collection has the expected number of elements..Should().BeEquivalentTo(dataTable)
: Performs a deep equivalency comparison between the subject and the expectation.
When comparing the rows of a DataTable
, by default the rows are matched by their index in the Rows
collection. But, if the subject and expectation tables contain equivalent PrimaryKey
values, then it is possible to match the rows by their primary key field data, irrespective of the row order within the Rows
collection. See "Equivalency Assertion Options" below for how to configure this.
The following assertions are available on DataColumn
objects:
.Should().BeEquivalentTo(dataColumn)
: Performs a deep equivalency comparison between the subject and the expectation.
The following assertions are available on DataRow
objects:
.Should().HaveColumn(columnName)
,.Should().HaveColumns(columnName, columnName, ...)
: Asserts that theDataTable
'sColumns
collection contains at least columns with the specified names. Additional columns are ignored..Should().BeEquivalentTo(dataRow)
: Performs a deep equivalency comparison between the subject and the expectation. This includes comparing field values, for which the defined columns must match.
When checking the equivalency of two DataRow
objects, by default the RowState
must match. But, if this is overridden using equivalency assertion options (.Excluding(row => row.RowState)
), two DataRow
objects with differing RowState
s can still be considered equivalent based on their field values. FluentAssertions automatically determines which version of field values to use in the subject and the expectation separately.
- For
DataRowState.Unchanged
, field values forDataRowVersion.Current
are used. - For
DataRowState.Added
, field values forDataRowVersion.Current
are used. - For
DataRowState.Deleted
, field values forDataRowVersion.Original
are used. - For
DataRowState.Modified
, field values forDataRowVersion.Current
are used.
In addition, if both the subject and the expectation are in the DataRowState.Modified
state, then the DataRowVersion.Original
values are also compared, separately from the DataRowVersion.Current
values. This can be disabled using the .ExcludingOriginalData()
equivalency assertion option.
Each DataSet
has a DataTableCollection
called Tables
, and each DataTable
has a DataColumnCollection
called Columns
and a DataRowCollection
called Rows
. Some assertions can be performed on these collection types.
The following assertions are in common to all three collection types:
.Should().BeEmpty()
: Succeeds if the collection contains no items (tables, columns, rows)..Should().NotBeEmpty()
: Succeeds if the collection contains at least one item (table, column, row)..Should().ContainEquivalentOf(x)
: Succeeds if the collection contains an item (table, column, row) that is equivalent to the supplied item..Should().NotContainEquivalentOf(x)
: Succeeds if the item does not contain any item (table, column, row) that is equivalent to the supplied item..Should().HaveSameCount(x)
: Succeeds if the collection contains the same number of items as the supplied collection of the same type..Should().NotHaveSameCount(x)
: Succeeds if the collection does not contain the same number of items as the supplied collection of the same type..Should().HaveCount(x)
: Succeeds if the collection contains exactly the specified number of items..Should().HaveCount(predicate)
: Succeeds if the predicate returns true for the number of items in the collection..Should().NotHaveCount(x)
: Succeeds if the collection contains a different number of items than the supplied count..Should().HaveCountGreaterThan(x)
: Succeeds if the collection contains more items than the supplied count..Should().HaveCountGreaterThanOrEqualTo(x)
: Succeeds if the collection contains at least as many items as the supplied count..Should().HaveCountLessThan(x)
: Succeeds if the collection contains fewer items than the supplied count..Should().HaveCountLessThanOrEqualTo(x)
: Succeeds if the collection contains at most as many items as the supplied count.
When checking equivalency, the operation can be fine-tuned by configuring the options provided to an optional configuration callback in the .BeEquivalentTo
method.
In addition to standard equivalency assertion options, which are applied recursively to all subjects found in the object graph, the following options are available that are specific to equivalency tests for supported System.Data
types:
.AllowingMismatchedTypes()
: Allows objects with equivalent data to be considered equivalent, even if their data types do not match. See "TypedDataSet
,DataTable
,DataRow
objects" below..IgnoringUnmatchedColumns()
: Allows tables to be equivalent if one table has columns that the other does not. The column definitions and row data for these columns is ignored..UsingRowMatchMode(Index | PrimaryKey)
: Specifies the manner in which rows in twoDataTable
s should be matched up for equivalency testing..ExcludingOriginalData()
: Specifies that when comparing twoDataRow
objects whoseRowState
is bothModified
, only theCurrent
value of fields should be compared, ignoring any differences in theOriginal
versions of those fields..Excluding(x => x.Member)
: Excludes the specified member of the subject type from comparison. This method can only be used to select members accessible from the subject type directly..ExcludingRelated((DataRelation x) => x.Member)
:.ExcludingRelated((DataTable x) => x.Member)
:.ExcludingRelated((DataColumn x) => x.Member)
:.ExcludingRelated((DataRow x) => x.Member)
:.ExcludingRelated((DataConstraint x) => x.Member)
: Excludes the specified member of a related System.Data type from comparison. This is to handle the case where a change to one object within aDataSet
causes corresponding changes automatically in other places. For instance, settingUnique
totrue
in aDataColumn
has a side-effect of adding a newUniqueConstraint
to theConstraints
collection of the containingDataTable
..ExcludingTable(tableName)
,.ExcludingTables(tableName, tableName, ...)
: Excludes tables with the specified name(s) from comparison..ExcludingColumnInAllTables(columnName)
,.ExcludingColumnsInAllTables(columnName, columnName, ...)
: Excludes all columns in any table within theDataSet
with the specified name(s) from comparison..ExcludingColumn(tableName, columnName)
,.ExcludingColumns(tableName, columnName, columnName, ...)
: Excludes the specified column(s) in tables with a specific name within theDataSet
s from comparison..ExcludingColumn(dataColumn)
,.ExcludingColumns(dataColumn, dataColumn, ...
): Excludes the specified column(s) in tables with names matching the ownerDataTable
of the suppliedDataColumn
object(s) from comparison. This method is supplied because with typedDataTable
objects, it is a common pattern that accessors forDataColumn
objects are provided.
These configuration methods follow a fluent initialization pattern; they return the same options object they were called on, so that configuration can be chained in a single expression.
See the "Examples" section for some uses of equivalency options.
If some of your DataSet
, DataTable
or DataRow
objects are typed objects (e.g. DataTable
classes that inherit from TypedTableBase<TRow>
), by default, equivalency tests will fail if the data types do not also match. If you want to compare typed objects with untyped objects that otherwise contain equivalent data, then the option AllowingMismatchedTypes
can be used:
DataSet expected = GetExpectedDataSet();
EmployeeData actual = service.GetEmployeeData();
actual.Should().BeEquivalentTo(expected, options => options.AllowingMismatchedTypes());
Apart from the default check for matching types, whether a DataSet
, DataTable
or DataRow
is a vanilla object or an instance of a custom subtype is irrelevant for the purposes of comparison.
Asserting that a DataTable
has certain columns:
var table = GetDataTable();
MyTypedTableType templateTable = GetExpectedDataTable();
table.Should().HaveColumn("FirstName");
table.Should().HaveColumns(templateTable.FirstNameColumn, templateTable.LastNameColumn);
Excluding tables and columns from equivalency tests:
You can exclude a column from consideration by name across all tables in a
DataSet
, or from a specificDataTable
.
var expected = GetExpectedDataSet();
var actual = GetActualDataSet();
actual.Should().BeEquivalentTo(expected, options => options
.ExcludingTable("Employees")
.ExcludingColumnInAllTables("EmployeeID")
.ExcludingColumn(tableName: "Employees", columnName: "Address"));
Excluding fields from types other than the subject:
DataSet
objects have a hierarchy of objects representing the data and many interrelations between these objects. For instance, settingUnique
totrue
in aDataColumn
has a side-effect of adding a newUniqueConstraint
to theConstraints
collection of the containingDataTable
. As such, to eliminate side-effects it may be necessary to ignore additional properties of types related to the subject. To support this, a methodExcludingRelated
can be used that allows members to be selected on any related type.
var expected = GetExpectedDataSet();
var actual = GetActualDataSet();
actual.Should().BeEquivalentTo(expected, options => options
.Excluding(dataSet => dataSet.DataSetName)
.ExcludingRelated((DataRelation relation) => relation.DataSet);
Matching the rows of DataTable
s by their primary key values:
When comparing the rows of a
DataTable
, by default, rows are expected to be occur at the same indices. If aDataTable
has aPrimaryKey
set, then it is possible to instead match rows by their primary key, irrespective of the order in which they occur.
var expected = GetExpectedDataTable();
var actual = GetActualDataTable();
actual.Should().BeEquivalentTo(expected, options => options.UsingRowMatchMode(RowMatchMode.PrimaryKey));
All credits for this extension go to Jonathan Gilbert.