compress_tc

Version 1.0.0 | 2026-04-08

compress_tc aggressively reduces memory use in string-heavy datasets by first converting fixed-width string variables to strL, then running Stata's built-in compress so short or unique strings can move back to ordinary storage when that is smaller. It is a fork of Luke Stein's strcompress with additional reporting, option control, and safer validation.

Requirements

• Stata 16 or later

Installation

capture ado uninstall compress_tc
net install compress_tc, from("https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/compress_tc") replace

Commands

Command	Description
compress_tc	Convert fixed-width string variables to strL, then optimize storage with compress

Quick Start

Use a built-in dataset first to see the command shape. auto has one string variable, make, so the example is simple but fully runnable.

sysuse auto, clear
compress_tc make, detail
display "Saved " r(bytes_saved) " bytes (" %4.1f r(pct_saved) "%)"

In a real string-heavy dataset, the savings are usually much larger than in sysuse auto.

How It Works

• Stage 1 converts the requested str# variables to strL unless you specify nostrl.
• Stage 2 runs compress unless you specify nocompress.
• detail shows the original string storage types before conversion, while varsavings reports the final per-variable type summary.
• noreport suppresses compress's detailed output but still shows the summary. quietly suppresses all output while preserving r().
• Memory reporting is dataset-wide because Stata's memory command reports dataset-wide usage.

Worked Examples

1. Compress every string variable in memory

If you omit a varlist, compress_tc scans the whole dataset and processes every fixed-width string variable it finds.

sysuse auto, clear
compress_tc

2. Inspect which variables changed

detail shows the original types before conversion. varsavings adds a per-variable summary after compression.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/prescriptions.dta", clear
compress_tc atc drug_name, detail varsavings

This is the most useful pattern when you want to understand where the memory savings are coming from.

3. Compare the two stages separately

Use nocompress to isolate the strL conversion, or nostrl to keep only ordinary compress behavior with the same reporting layer.

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/procedures.dta", clear
compress_tc kva_code proc_description, nocompress

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/procedures.dta", clear
compress_tc kva_code proc_description, nostrl

4. Run quietly inside a larger workflow

quietly suppresses console output but still leaves the summary results in r().

use "https://raw.githubusercontent.com/tpcopeland/Stata-Tools/main/_data/prescriptions.dta", clear
compress_tc, quietly
return list

Key Options

Option	What it does
nocompress	Skip the final compress step and keep only the strL conversion
nostrl	Skip the strL conversion and run ordinary compress only
noreport	Suppress compress's per-variable output while keeping the summary
quietly	Suppress all output while still returning results in r()
detail	Show each processed string variable's original storage type
varsavings	Show a per-variable summary after compression

Returned Results

compress_tc stores the following in r():

• r(bytes_saved): total bytes saved
• r(pct_saved): percentage reduction in data size
• r(bytes_initial): initial data size in bytes
• r(bytes_final): final data size in bytes
• r(varlist): string variables actually processed

Technical Notes

• strL storage is especially useful for repeated values, long text, and sparse strings.
• For variables with short, unique strings, strL can temporarily increase memory use. The second-stage compress call is what re-optimizes those cases.
• Datasets that contain strL variables must be saved in Stata 13+ .dta format.

Version History

• 1.0.0 (2026-04-08): Initial Stata-Tools release of the two-stage string-compression workflow.

Author

Timothy P Copeland, Karolinska Institutet

Fork of strcompress by Luke Stein.