Moving from SAS, SSRS or Crystal
Most teams that adopt DRE already have reports somewhere else: SQL Server Reporting Services
(SSRS), SAS programs, Power BI paginated reports, Crystal Reports, or a SQL script and a spreadsheet someone
fills in by hand. The dre-port agent skill moves them over with you, one
report at a time. It checks the report specification and, when an old output is available,
compares keyed records and totals for matching inputs, parameters and dates. Guided SAS support
is included in skills 5.1; SAS execution is not required.
DRE has no converter command. Your coding agent reads the old report, translates it with the skill’s mapping reference, asks you about anything the files don’t say, and writes ordinary DRE files you can read and change.
What it reads
Section titled “What it reads”| You have | What the agent reads |
|---|---|
SAS .sas source and referenced files |
programs, includes, macro definitions, inputs, parameters and formats; missing dependencies and unsupported constructs remain explicit blockers |
| a report specification without source or old output | business rules and supplied schemas; results are business-rule validation, not verified SAS equivalence |
SSRS, Power BI paginated or Report Builder .rdl / .rdlc files |
datasets (query text or stored procedure), parameters, groups, expressions, page header and footer |
| SSRS subscriptions | a csv you export from the report server’s catalog database (the skill has the read-only query) |
| Crystal Reports | an XML export of the report definition, or a “Report Definition” text export. Not a binary .rpt: Crystal’s file format needs SAP’s SDK, so export it first |
| anything else (Jasper, BI Publisher, Access, an Excel macro) | the SQL, a screenshot or PDF of the output, its parameters, and who gets it when; the agent asks for the rest |
Where your data lives
Section titled “Where your data lives”DRE reads DuckDB, Postgres, Databricks, Snowflake, BigQuery and ClickHouse. Its optional JDBC source also connects Oracle and SQL Server with a user-supplied vendor JAR and operator-maintained Java 17+. Preparation runs inside the source database. The agent uses authorized live catalog metadata or supplied schema definitions, and marks unknown names, types and relationships as unresolved. It explains generated statements and obtains review before executing data-changing SQL, including temporary preparation.
SAS scope and worked example
Section titled “SAS scope and worked example”The bounded reporting path covers PROC SQL selection/joins/aggregation, simple row-local filters and derived columns, scalar parameters, deterministic sorting/deduplication, basic summaries and tabular output. The agent reviews missing values, dates, ordering, merge cardinality and rounding explicitly. Stateful DATA steps, complex macro execution, local SAS datasets and specialist procedures need supplied business rules or remain unsupported; DRE is not a SAS interpreter or a replacement for the full SAS suite.
The synthetic overdue-loan project shows an unexecuted SAS reference beside runnable Oracle and SQL Server variants, independent expected results and both ordinary and template-designed workbooks. Its checks establish DRE business-rule results, not executed SAS parity. Use it without a SAS installation or company data. Layout changes and unsupported logic are part of the migration assessment, not silently accepted equivalence.
How a report maps
Section titled “How a report maps”| Old report | DRE |
|---|---|
| a dataset’s query | a .sql file in the report folder |
| a stored procedure | its logic as SQL (temp tables in a setup file) |
| parameters, with defaults and available values | variables with var_schema; “last month” defaults come from run.date |
| a table or list | an xlsx tab, or a table in a PDF template |
| groups with subtotals | ROLLUP in SQL, or a PDF template with a heading and subtotal per group |
expressions and formulas (IIf, Format, Crystal {@formulas}) |
SQL columns, and column formats in the output |
| page header, footer, “Page N of M” | the PDF template’s @page rules |
| “no rows” message | on_empty |
| email and file share subscriptions, schedules | destinations and schedules.yml |
| data-driven subscriptions, bursting per customer | per-row outputs (for_each) |
| one copy of a report per region or client | one report with Sets |
| interactive sort, drill-through, toggles, charts | no direct equivalent: DRE makes files. The agent asks what to do instead |
Validation and optional parity
Section titled “Validation and optional parity”When an old output is available, export the old report for one set of parameter values (the last complete
month, say) as csv or xlsx. The agent runs the DRE report with the same values, writing its rows
with dre run <report> --no-deliver --output-json parity, and compares the two with a small
script bundled with the skill. It reports:
- row counts on each side;
- column totals for every numeric column;
- a cell diff keyed on the columns that identify a row, each difference explained as a real value difference, rounding (within a tolerance), formatting (the same date or number written another way) or order;
- rows only on one side, and duplicate keys.
With no reference output, the agent instead checks explicit expected keyed records, boundary dates, classifications, counts, totals and rounding from the business specification. It reports expected versus actual values and labels SAS equivalence unverified. Missing dependencies and unresolved semantics keep a migration partial, even if its DRE configuration validates.
Real differences are fixed (or one of the two reports is shown to be wrong) and the check runs again, until it matches. For a PDF or a styled workbook you also compare the old and new files side by side.
A whole estate
Section titled “A whole estate”Pointed at a folder of .rdl files and Crystal exports, the skill first lists every report: its
sources, parameters, subscriptions and an effort estimate (S to XL, with the reasons, such as
stored procedures, custom code or charts). It groups near-duplicates (a copy per region) as
candidates for one report with Sets, and suggests an order, simplest first. Then reports are
ported one at a time. SAS estates include a source/dependency inventory by reading the programs;
the bundled XML inventory script does not parse SAS.
Practices
Section titled “Practices”The skill follows the porting rules in DRE practices: logic in SQL, not templates (PORT-1); explicit validation evidence before switching over (PORT-2); one report at a time, simplest first (PORT-3); first deliveries held or on dev (PORT-4); a record of where each report came from (PORT-5); and dependency/semantic review before a complete-port claim (PORT-6).
Try it
Section titled “Try it”Install the agent skills, then ask your agent something like “port this
SSRS report to DRE” with the .rdl file in your project, or “list the reports in ./ssrs-export
and suggest where to start”.