How-to guide

Short recipes for everyday work. Each names the key or command in the editor and the dsr command that does the same in a terminal of the editor (where an AI agent works too). The manual explains each tool in full.

Run only what I changed

After one full run the objects stay in the session. Select the lines and press Cmd/Ctrl+Enter, or run a range:

dsr run prog.R --from-line 12 --to-line 30

A range runs whole expressions. Run to Line and Run from Line (right-click in the editor) do the same from the cursor.

Find out why a variable has a wrong value

Stop the run where the value is made, for the record you care about:

dsr run prog.R --stop-at 'R/derive.R:5 if any(bmi == 30)'
dsr pry eval 'bmi'
dsr pry next
dsr pry eval 'grp[bmi == 30]'
dsr pry quit

In the editor: a breakpoint in the margin with a condition (right-click the red dot), then F5; or the cursor on the line and Stop at Line (Cmd/Ctrl+Alt+F9). The stop works inside a function too.

Look inside a run that failed

An ERROR keeps the frames where it happened:

dsr pry stack
dsr pry ls
dsr pry eval 'summary(weight)'

The editor opens the same frames in the Pry panel.

Stop a loop at one subject

Write the stop into the program and run it as always:

for (i in seq_len(nrow(dm))) {
  pry(when = dm$USUBJID[i] == "S-110")
  ...
}

pry(stop = FALSE) keeps a snapshot and goes on; dsr pry points lists the snapshots.

See which pipeline step loses records

Put the cursor on the pipeline and run Debug the Pipeline at the Cursor, or:

dsr pipe programs/adam/ad_adsl.R:42

Each step shows its rows before and after, the columns it added or dropped, and the records it dropped (by USUBJID and the sequence number). dsr data '@pipe[1]:dropped' lists them.

Find which line of a function made a value

dsr trace --call 'bmi_group(c(24, 30, 31))'

Every line that ran, the variables it changed, each if with the values it compared. In the editor: Trace Function Call, or Call a Function with a Form to fill in the arguments.

Compare my data set with a second programmer's

dsr compare adsl adsl_ref --key USUBJID

As PROC COMPARE: the rows and columns only in one, and each value that differs, with examples. In the editor: Compare Two Data Sets. A file works too: .rds, .xpt, .sas7bdat, .csv.

Check a table against the CSR

dsr numbers output/t14_3_01.txt csr.txt --table 14-3.01

Row by row and number by number at the reference's decimals: NUMBERS: MATCH or each difference with its line. --round sas when the reference was rounded by SAS; --page 49 for a page of a PDF.

Count subjects, not records

dsr freq adae AEDECOD --subjects

--by TRT01A gives percents within each arm; dsr means adsl AGE --by TRT01A gives n, mean, SD, min, median, max.

Find the mistakes before running

Save the file: lint marks functions that do not exist in your installed packages, argument names a function does not have, and names with a typo, each with "did you mean". Or for a folder:

dsr lint programs

Tidy a program without changing what it does

Format Document (Shift+Alt+F), or:

dsr format prog.R

The file is written only when its tokens and comments are the same before and after; --check writes nothing.

Prove a program runs on its own

dsr batch prog.R

A new R session for each program, as a production run, and prog.log next to it. --vanilla leaves out the project's .Rprofile.

Let Claude Code work in my session

DataStep R: Set Up Agent Tools writes the MCP configuration of the project. Or in a terminal of the editor, where dsr is on the PATH, tell the agent to read dsr guide once. Its runs and stops show in your R Console and editor. The other agents: the page Set up an agent.