Five compression algorithms for embedded sensor data. Zero allocations. Zero dependencies. #include and go.

microcodec is a small C99 compression library focused on the data patterns that actually show up in embedded and IoT systems:

• repeated bytes and zero-padded frames
• small counters, IDs, and flags
• slowly changing sensor readings
• monotonic timestamps
• compact binary payloads

The library is designed around caller-owned buffers, no heap usage, no hidden state, and simple APIs that fit microcontroller codebases.

Additional project resources:

• Changelog
• Contributing guide
• Security policy
• Documentation index
• Wiki source pages

#include "microcodec.h"

float temperatures[64];
uint8_t encoded[512];
float recovered[64];

mc_buf_t dst = MC_BUF(encoded, sizeof(encoded));

if (mc_delta_encode_f32(temperatures, 64, 1, &dst) == MC_OK) {
    mc_delta_decode_f32(MC_SLICE(encoded, dst.len), recovered, 64);
}

For general binary blobs:

mc_lzss_ctx_t ctx;
mc_lzss_ctx_init(&ctx);

uint8_t out[mc_lzss_max_encoded_size(256)];
mc_buf_t dst = MC_BUF(out, sizeof(out));

mc_err_t err = mc_lzss_encode(&ctx, MC_SLICE(data, data_len), &dst);
if (err == MC_ERR_INCOMPRESS) {
    /* store raw data instead */
}

All configuration is compile-time and lives in microcodec.h.

#define MICROCODEC_ENABLE_RLE    1
#define MICROCODEC_ENABLE_VARINT 1
#define MICROCODEC_ENABLE_DELTA  1
#define MICROCODEC_ENABLE_LZSS   1
#define MICROCODEC_ENABLE_HUFF   1

#define MICROCODEC_LZSS_WINDOW_SIZE   256
#define MICROCODEC_LZSS_MIN_MATCH     3
#define MICROCODEC_LZSS_MAX_MATCH     18

#define MICROCODEC_HUFF_SYMBOLS       256
#define MICROCODEC_HUFF_MAX_CODE_LEN  15

#define MICROCODEC_DELTA_F32_SCALE    100.0f

Main families:

• mc_rle_encode / mc_rle_decode
• mc_varint_encode_* / mc_varint_decode_*
• mc_delta_encode_* / mc_delta_decode_*
• mc_lzss_ctx_init, mc_lzss_encode, mc_lzss_decode
• mc_huff_build_table, mc_huff_encode_static, mc_huff_decode_static
• mc_huff_encode_adaptive, mc_huff_decode_adaptive

Dispatch API:

• mc_encode
• mc_decode
• mc_max_encoded_size
• mc_alg_name

Current dispatch contract in this repo:

• MC_ALG_RLE, MC_ALG_LZSS, MC_ALG_HUFF operate on raw bytes
• MC_ALG_VARINT operates on uint32_t[] passed via src.ptr
• MC_ALG_DELTA operates on float[] passed via src.ptr with default order 1

microcodec is intentionally separate from microdb. A typical pattern is:

float readings[50];
uint8_t compressed[256];

mc_buf_t cbuf = MC_BUF(compressed, sizeof(compressed));
mc_delta_encode_f32(readings, 50, 1, &cbuf);

microdb_kv_put(&db, "sensor_burst", compressed, cbuf.len);

On load, read the compressed bytes from microdb and decode them back into the target array with the matching mc_*_decode call.

• No heap allocation anywhere in the codec implementations
• No global mutable state
• All APIs are caller-buffer driven
• Compile-time feature toggles for each algorithm
• Separate focused codecs instead of one opaque auto-tuned compressor

microcodec is intended to live alongside small focused embedded libraries such as microdb and other components in the micro-toolkit ecosystem.

MIT