run-sas

name	run-sas
description	Run SAS programs via connect_saspy, parse HTML output into DataFrames, and work with results. Use when asked to run, execute, or submit SAS code, or read/analyze SAS output.

Running SAS Programs with connect_saspy

Patterns for submitting SAS code and reading results using connect_saspy, pandas, and BeautifulSoup.

---

Step 1 — Ensure Required Packages Are Installed

Before any SAS work, verify and install all dependencies in a single idempotent block.

import subprocess, sys

def _pip_install(*packages):
    subprocess.check_call([sys.executable, "-m", "pip", "install", "--quiet", *packages])

try:
    import saspy
except ImportError:
    _pip_install("saspy")
    import saspy

try:
    import connect_saspy
except ImportError:
    _pip_install("git+https://github.com/matise-joe-norc/connect-saspy.git")
    import connect_saspy

try:
    import pandas as pd
except ImportError:
    _pip_install("pandas")
    import pandas as pd

try:
    import numpy as np
except ImportError:
    _pip_install("numpy")
    import numpy as np

try:
    from bs4 import BeautifulSoup
except ImportError:
    _pip_install("beautifulsoup4", "lxml")
    from bs4 import BeautifulSoup

Required packages summary:

Package	Source	Purpose
saspy	PyPI	SAS session backend used by connect_saspy
connect_saspy	GitHub (matise-joe-norc)	Simplified SAS connection wrapper
pandas	PyPI	Parse HTML tables from LST output; data wrangling
numpy	PyPI	Numerical operations on results
beautifulsoup4	PyPI	Fine-grained HTML parsing of LST output
lxml	PyPI	Fast HTML parser backend required by pandas.read_html

---

Step 2 — Submit SAS Code

connect_saspy auto-starts a SAS session on first import and exposes a module-level submit().
The session uses results="html" — all LST output is HTML.

Module-level convenience (recommended for simple use)

import connect_saspy

lst = connect_saspy.submit("""
    proc print data=sashelp.class (obs=5);
    run;
""")
# lst is an HTML string, or None if SAS produced no printed output
print(lst)

Class-level (recommended when managing multiple connections or lifecycle)

from connect_saspy import SASConnection

sas = SASConnection()

lst = sas.submit("""
    data work.example;
        set sashelp.cars (obs=10);
    run;

    proc means data=work.example;
        var MSRP Invoice;
    run;
""")

sas.close()

What submit() returns:

• The SAS LST (list/output) as an HTML string, or None if no output was produced.
• The SAS LOG is always printed to stdout automatically — watch it for ERROR: and WARNING: lines.

---

Step 3 — Parse Output into DataFrames

Because the LST is HTML, use pandas.read_html() to extract tables in one line.

import pandas as pd
from io import StringIO

def sas_lst_to_dataframes(lst: str | None) -> list[pd.DataFrame]:
    """Return all HTML tables in the LST as a list of DataFrames."""
    if not lst:
        return []
    return pd.read_html(StringIO(lst))

# Example
lst = connect_saspy.submit("proc print data=sashelp.class; run;")
dfs = sas_lst_to_dataframes(lst)

for i, df in enumerate(dfs):
    print(f"--- Table {i} ---")
    print(df.head())

Fine-grained parsing with BeautifulSoup

Use BeautifulSoup when you need to inspect titles, footnotes, or non-table content.

from bs4 import BeautifulSoup

def parse_sas_output(lst: str | None) -> dict:
    """Extract titles, tables, and footnotes from SAS HTML output."""
    if not lst:
        return {"titles": [], "tables": [], "footnotes": []}

    soup = BeautifulSoup(lst, "lxml")
    return {
        "titles":     [t.get_text(strip=True) for t in soup.find_all(class_="c")],
        "tables":     pd.read_html(StringIO(lst)) if soup.find("table") else [],
        "footnotes":  [f.get_text(strip=True) for f in soup.find_all(class_="f")],
    }

result = parse_sas_output(lst)
print(result["titles"])
df = result["tables"][0]  # first table as DataFrame

---

Step 4 — Check for Errors in the LOG

The LOG is printed to stdout by submit(), but you can also inspect it programmatically via the underlying saspy session when you need to detect errors in code.

from connect_saspy import get_connection

def submit_and_check(code: str) -> tuple[str | None, bool]:
    """Submit SAS code; return (lst, had_errors)."""
    sas_conn = get_connection()
    # Access the raw saspy session for log inspection
    session = sas_conn._sas_session
    if session is None:
        sas_conn._start_connection()
        session = sas_conn._sas_session

    result = session.submit(code)
    log  = result.get("LOG", "")
    lst  = result.get("LST")

    had_errors = any(
        line.lstrip().startswith(("ERROR:", "WARNING:"))
        for line in log.splitlines()
    )

    print("=== SAS LOG ===")
    print(log)
    print("=== END LOG ===")

    return lst, had_errors

lst, had_errors = submit_and_check("proc print data=sashelp.class; run;")
if had_errors:
    print("SAS reported errors or warnings — review the log above.")

---

Common Patterns

Read a SAS dataset into a pandas DataFrame

lst = connect_saspy.submit("""
    proc print data=sashelp.cars noobs;
    run;
""")
df = sas_lst_to_dataframes(lst)[0]

Run PROC MEANS and capture statistics

lst = connect_saspy.submit("""
    proc means data=sashelp.cars n mean std min max;
        var MSRP Invoice;
    run;
""")
stats_df = sas_lst_to_dataframes(lst)[0]
print(stats_df)

Pass Python data to SAS via a data step

import pandas as pd

# Build a SAS data step from a Python dict
rows = [("North", 1000), ("South", 1500), ("East", 1200), ("West", 800)]
datalines = "\n".join(f"{region} {sales}" for region, sales in rows)

code = f"""
data work.sales;
    input region $ sales;
    datalines;
{datalines}
    ;
run;

proc means data=work.sales;
    var sales;
run;
"""

lst = connect_saspy.submit(code)
dfs = sas_lst_to_dataframes(lst)

Close the connection explicitly

# Module-level global connection
connect_saspy.close_connection()

# Class-instance connection
sas = connect_saspy.SASConnection()
# ... work ...
sas.close()

---

Troubleshooting

Symptom	Likely Cause	Fix
SASConfigNotFoundError	sascfg_personal.py missing or misconfigured	See SASPy configuration docs
lst is None	SAS code ran but produced no printed output	Normal for DATA steps; use PROC PRINT to see data
ERROR: in LOG	SAS syntax error or missing dataset	Read the LOG carefully; correct the SAS code
pd.read_html raises ValueError	LST has no <table> elements	Use BeautifulSoup to inspect raw HTML first
Import fails for connect_saspy	Package not installed from GitHub	Run pip install git+https://github.com/matise-joe-norc/connect-saspy.git

---

saspy Configuration Prerequisite

connect_saspy delegates all SAS connectivity to saspy, which requires a sascfg_personal.py configuration file. Common access methods:

# Local SAS installation (IOM)
# sascfg_personal.py example:
SAS_config_names = ['winlocal']
winlocal = {
    'saspath': '/path/to/sas',
}

# SAS Viya / SAS Studio compute
# See: https://sassoftware.github.io/saspy/install.html

When saspy is not yet configured, run import saspy; saspy.SAScfg to find where the config file should be placed, then follow the SASPy install guide.