Source

The 1. Source tab defines where rows are read from. Every row of the source runs through the steps once.

Insert query for a table… lists the tables (or views) of the connection and writes a query that reads all their columns, as a starting point. Load fields reads the field list of the query, without reading its rows, and fills Source fields with the name, type and display name of every field the query returns. Load the fields before mapping: the field panel of the Steps tab, Map same names and the validation of formulas use them.

Dataverse: FetchXML

Write the query as FetchXML, for example in FetchXML Builder (another XrmToolBox tool) and paste it here:

<fetch>
  <entity name="contact">
    <attribute name="fullname" />
    <attribute name="emailaddress1" />
    <attribute name="parentcustomerid" />
    <attribute name="statecode" />
    <filter>
      <condition attribute="statecode" operator="eq" value="0" />
    </filter>
    <link-entity name="account" from="accountid" to="parentcustomerid" link-type="outer" alias="acc">
      <attribute name="accountnumber" />
    </link-entity>
  </entity>
</fetch>

How the result becomes source fields:

• Every attribute is a field with its logical name: fullname, emailaddress1.
• Lookups have the value table:guid (for example account:b9564ad3-…), which can be written to another lookup column as it is. A second field <name>name holds the display name of the referenced record: parentcustomeridname.
• Choices, yes/no columns and multi-select choices have their numeric value (1;3 for multi-select). The <name>name field holds the label in the user’s language: statecodename.
• Linked tables are prefixed with their alias: acc.accountnumber. Give every link-entity an explicit alias so the names stay stable; without one Replisteer uses <table><n>, for example account1.
• Columns with an alias attribute (including aggregates) use the alias as their name.
• The primary key of the main table is always available as <table>id, for example contactid, even when it is not selected.

Paging. Replisteer reads all records page by page. The page size is the count attribute of the FetchXML when present, otherwise the page size in Options… (500 by default), and at most 5,000. Queries with top or aggregate="true" are read in a single request.

SQL Server: SQL query

Any query that returns rows works: a SELECT, a view, or a stored procedure call (EXEC dbo.GetChanges).

SELECT c.Id, c.FirstName, c.LastName, c.Email, a.AccountNumber
FROM dbo.Contacts AS c
LEFT JOIN dbo.Accounts AS a ON a.Id = c.AccountId
WHERE c.ModifiedOn >= DATEADD(day, -1, SYSUTCDATETIME())

Every column of the result is a field with its column name; give computed columns an alias (AS FullName). Load fields asks SQL Server for the columns without running the query; procedures that build their result in temporary tables sometimes cannot answer that. The run itself still works, map the columns by name. Values keep their SQL types: numbers stay numbers, dates stay dates. The command timeout of the connection (30 seconds by default) applies to the query.

CSV file

The whole file is read; there is no query. With a header row the fields are named after the header (blank names become Column3, repeated names get _2, _3). Without a header they are Column1, Column2, and so on. Every value is text; see CSV files for how text that holds numbers and dates behaves in formulas.