From a83ecdff2f0eb8d6ad4798544f5249a7ea461607 Mon Sep 17 00:00:00 2001
From: Aaron Kimbrell <aronwk.aaron@gmail.com>
Date: Sun, 27 Sep 2026 10:05:10 -0500
Subject: [PATCH] feat(common): streaming raw deflate and CRC-32 in
 ZCompression

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 dCommon/ZCompression.cpp | 58 ++++++++++++++++++++++++++++++++++++++++
 dCommon/ZCompression.h   | 33 ++++++++++++++++++++++-
 2 files changed, 90 insertions(+), 1 deletion(-)

diff --git a/dCommon/ZCompression.cpp b/dCommon/ZCompression.cpp
index 344ac80a2..5f02bf63c 100644
--- a/dCommon/ZCompression.cpp
+++ b/dCommon/ZCompression.cpp
@@ -1,5 +1,7 @@
 #include "ZCompression.h"
 
+#include <algorithm>
+
 #include "zlib.h"
 
 namespace ZCompression {
@@ -97,4 +99,60 @@ namespace ZCompression {
 		if (!out || out->size() != size) return std::nullopt;
 		return out;
 	}
+
+	uint32_t Crc32(uint32_t crc, std::string_view data) {
+		while (!data.empty()) {
+			const auto piece = static_cast<uInt>(std::min<size_t>(data.size(), 1u << 30));
+			crc = static_cast<uint32_t>(crc32(crc, reinterpret_cast<const Bytef*>(data.data()), piece));
+			data.remove_prefix(piece);
+		}
+		return crc;
+	}
+
+	struct RawDeflater::State {
+		z_stream stream{};
+		bool ok{};
+	};
+
+	RawDeflater::RawDeflater(Sink sink, int level) : m_State(std::make_unique<State>()), m_Sink(std::move(sink)) {
+		m_State->ok = deflateInit2(&m_State->stream, level, Z_DEFLATED, -15, 8, Z_DEFAULT_STRATEGY) == Z_OK;
+	}
+
+	RawDeflater::~RawDeflater() {
+		deflateEnd(&m_State->stream);
+	}
+
+	bool RawDeflater::Run(std::string_view data, int flush) {
+		if (!m_State->ok) return false;
+		auto& stream = m_State->stream;
+		stream.next_in = reinterpret_cast<Bytef*>(const_cast<char*>(data.data()));
+		stream.avail_in = static_cast<uInt>(data.size());
+		char buffer[64 * 1024];
+		int result = Z_OK;
+		do {
+			stream.next_out = reinterpret_cast<Bytef*>(buffer);
+			stream.avail_out = sizeof(buffer);
+			result = deflate(&stream, flush);
+			if (result == Z_STREAM_ERROR) return m_State->ok = false;
+			const size_t made = sizeof(buffer) - stream.avail_out;
+			m_Out += made;
+			if (made && !m_Sink(std::string_view(buffer, made))) return m_State->ok = false;
+		} while (stream.avail_out == 0 || (flush == Z_FINISH && result != Z_STREAM_END));
+		return true;
+	}
+
+	bool RawDeflater::Write(std::string_view data) {
+		m_In += data.size();
+		while (data.size() > (1u << 30)) {
+			if (!Run(data.substr(0, 1u << 30), Z_NO_FLUSH)) return false;
+			data.remove_prefix(1u << 30);
+		}
+		return Run(data, Z_NO_FLUSH);
+	}
+
+	bool RawDeflater::Finish() {
+		const bool ok = Run({}, Z_FINISH);
+		m_State->ok = false;
+		return ok;
+	}
 }
diff --git a/dCommon/ZCompression.h b/dCommon/ZCompression.h
index 57d399571..dd310b680 100644
--- a/dCommon/ZCompression.h
+++ b/dCommon/ZCompression.h
@@ -1,6 +1,8 @@
 #pragma once
 
 #include <cstdint>
+#include <functional>
+#include <memory>
 #include <optional>
 #include <string>
 #include <string_view>
@@ -20,5 +22,34 @@ namespace ZCompression {
 
 	// Raw deflate data (no zlib or gzip header, as in zip files) that inflates to `size` bytes; nullopt when it doesn't
 	std::optional<std::string> InflateRaw(std::string_view data, size_t size);
-}
 
+	// CRC-32 (as in zip and gzip files) of `data`, continuing from `crc` (0 to start)
+	uint32_t Crc32(uint32_t crc, std::string_view data);
+
+	/**
+	 * Raw deflate (as in zip files) of data given a piece at a time, so a big file never has to sit in memory: each
+	 * Write and Finish hands what's compressed so far to `sink`.
+	 */
+	class RawDeflater {
+	public:
+		using Sink = std::function<bool(std::string_view)>; // false stops (e.g. the disk is full)
+		explicit RawDeflater(Sink sink, int level = 6);
+		~RawDeflater();
+		RawDeflater(const RawDeflater&) = delete;
+		RawDeflater& operator=(const RawDeflater&) = delete;
+
+		bool Write(std::string_view data);
+		// The rest of the compressed data; the deflater can't be written to afterwards
+		bool Finish();
+		uint64_t BytesIn() const { return m_In; }
+		uint64_t BytesOut() const { return m_Out; }
+
+	private:
+		bool Run(std::string_view data, int flush);
+		struct State;
+		std::unique_ptr<State> m_State;
+		Sink m_Sink;
+		uint64_t m_In{};
+		uint64_t m_Out{};
+	};
+}