CoralCache：一個提高微服務可用性的中間件

摘要：當數據庫出問題時能降級從本地緩存的數據中查詢數據， CoralCache就是這樣一個提高微服務可用性的中間件。

背景

有些場景下，微服務依賴數據庫中一些配置項或者數量很少的數據，但當數據庫本身有問題時候，即使數據量很少，這個服務是不能正常工作；因此需要考慮一種能支持全量+極少變更的全局數據的場景，當數據庫出問題時能降級從本地緩存的數據中查詢數據，CoralCache就是這樣一個提高微服務可用性的中間件。

架構

CoralCache中間件架構如下圖所示，通過@EnableLocal註解開啓功能，應用啓動後將配置的表數據一次性加載到內存中，內存中的數據邏輯結構和數據庫中的邏輯結構一樣。

圖1. 架構圖

表達式計算引擎

內存查詢引擎的原理是數據庫查詢降級發生後，Intercepter將攔截到的原始SQL傳入查詢引擎中，查詢引擎解析SQL後得到表名、列名、where條件表達式，遍歷InnerDB中對應表的數據行，並通過表達式計算引擎計算結果，計算結果爲真則添加到結果集中最後返回給調用方。

計算引擎結構如下圖所示，將where條件表達式轉爲後綴表達式後依次遍歷後綴表達式，遇到操作數直接入棧，遇到操作符則根據操作符需要的操作數個數彈棧。

圖2. 表達式計算引擎結構

然後根據操作符和彈出的操作數進行計算，不同操作符對應不同的計算方法，並將計算後的結果重新作爲操作數入棧執到遍歷完成，核心計算流程代碼如下所示：

public Object calc(Expression where, InnerTable table, InnerRow row) {
        try {
            postTraversal(where);
        } catch (Exception e) {
            log.warn("calc error: {}", e.getMessage());
            return false;
        }
        for (ExprObj obj : exprList) {
            switch (obj.exprType()) {
                case ITEM:
                    stack.push(obj);
                    break;
                case BINARY_OP: {
                    ExprObj result = calcBinaryOperation(((ExprOperation) obj).getOperationType(), table, row);
                    stack.push(result);
                    break;
                }
                case UNARY_OP: {
                    ExprObj result = calcSingleOperation(((ExprOperation) obj).getOperationType(), table, row);
                    stack.push(result);
                    break;
                }
                case FUNCTION_OP: {
                    ExprObj result = calcFunctionOperation(((ExprOperation) obj).getOperationType(), table, row);
                    stack.push(result);
                    break;
                }
                default:
                    break;
            }
        }
        return stack.pop();
    }

常見運算符的實現

邏輯運算

邏輯常見運算符爲<、<=、>、>=、=等，它們的共性都是需要2個操作數並且返回值是布爾類型。

public ExprItem logicalCalculus(InnerTable table, InnerRow row, LogicalOperation logicalOperation) {

        ExprObj second = stack.pop();
        ExprObj first = stack.pop();

        ExprItem result = new ExprItem();
        result.setItemType(ItemType.T_CONST_OBJ);
        Obj firstObj = getObj((ExprItem) first, table, row);
        Obj secondObj = getObj((ExprItem) second, table, row);
        boolean value = logicalOperation.apply(firstObj, secondObj);
        result.setValue(new Obj(value, ObjType.BOOL));
        return result;
    }

例子，以"="的實現來展示：

private ExprObj calcBinaryOperation(OperationType type, InnerTable table, InnerRow row) {
        ExprObj result = null;
        switch (type) {
            case T_OP_EQ:
                result = logicalCalculus(table, row, (a, b) -> ObjUtil.eq(a, b)); // 等於符號的實現
                break;
            ...
            default:
                break;
        }
        return result;
 }

public class ObjUtil {
    private static ObjType resultType(ObjType first, ObjType second) {
        return ObjType.RESULT_TYPE[first.ordinal()][second.ordinal()];
    }

    public static boolean eq(Obj first, Obj second) {
        ObjType type = resultType(first.getType(), second.getType());

        switch (type) {
            case LONG: {
                long firstValue = first.getValueAsLong();
                long secondValue = second.getValueAsLong();
                return firstValue == secondValue;
            }
            case DOUBLE: {
                double firstValue = first.getValueAsDouble();
                double secondValue = second.getValueAsDouble();
                return Double.compare(firstValue, secondValue) == 0;
            }
            case TIMESTAMP: {
                java.util.Date firstValue = first.getValueAsDate();
                java.util.Date secondValue = first.getValueAsDate();
                return firstValue.compareTo(secondValue) == 0;
            }
            ...
            default:
                break;
        }
        throw new UnsupportedOperationException(first.getType() + " and " + second.getType() + " not support '=' operation.");
    }
}

數學運算

數學運算和邏輯運算的流程都一樣，只不過運算後的結果爲數字類型。

LIKE運算符

除了上面說的邏輯運算和數學運算外，還支持進行模糊匹配的特殊操作符LIKE。

LIKE表達式語法

常見用法如下

LIKE "%HUAWEI" 匹配以HUAWEI結尾的字符串
LIKE "HUAWEI%" 匹配以HUAWEI開頭的字符串
LIKE "A_B" 匹配以"A"起頭且以"Z"爲結尾的字串
LIKE "A?B" 同上
LIKE "%[0-9]%" 匹配含有數字的字符串
LIKE "%[a-z]%" 匹配含有小寫字母字符串
LIKE "%[!0-9]%"匹配不含數字的字符串
?和_都表示單個字符

JAVA中實現LIKE的方案：將LIKE的模式轉爲JAVA中的正則表達式。

LIKE詞法定義

expr := wild-card + expr
      | wild-char + expr
      | escape + expr
      | string + expr
      | ""

wild-card := %  
wild-char := _  
escape := [%|_]  
string := [^%_]+ (One or > more characters that are not wild-card or wild-char)

定義Token類

public abstract class Token {
    private final String value;

    public Token(String value) {
        this.value = value;
    }

    public abstract String convert();

    public String getValue() {
        return value;
    }
}

public class ConstantToken extends Token {
    public ConstantToken(String value) {
        super(value);
    }

    @Override
    public String convert() {
        return getValue();
    }
}

public class EscapeToken extends Token {
    public EscapeToken(String value) {
        super(value);
    }

    @Override
    public String convert() {
        return getValue();
    }
}

public class StringToken extends Token {
    public StringToken(String value) {
        super(value);
    }

    @Override
    public String convert() {
        return Pattern.quote(getValue());
    }
}

public class WildcardToken extends Token {
    public WildcardToken(String value) {
        super(value);
    }

    @Override
    public String convert() {
        return ".*";
    }
}

public class WildcharToken extends Token {
    public WildcharToken(String value) {
        super(value);
    }

    @Override
    public String convert() {
        return ".";
    }
}

創建Lexer(Tokenizer)

public class Tokenizer {

    private Collection<Tuple> patterns = new LinkedList<>();

    public <T extends Token> Tokenizer add(String regex, Function<String, Token> creator) {
        this.patterns.add(new Tuple<Pattern, Function<String, Token>>(Pattern.compile(regex), creator));
        return this;
    }

    public Collection<Token> tokenize(String clause) throws RuntimeException {
        Collection<Token> tokens = new ArrayList<>();
        String copy = String.copyValueOf(clause.toCharArray());

        int position = 0;
        while (!copy.equals("")) {
            boolean found = false;
            for (Tuple tuple : this.patterns) {
                Pattern pattern = (Pattern) tuple.getFirst();
                Matcher m = pattern.matcher(copy);
                if (m.find()) {
                    found = true;
                    String token = m.group(1);
                    Function<String, Token> fn = (Function<String, Token>) tuple.getSecond();
                    tokens.add(fn.apply(token));
                    copy = m.replaceFirst("");
                    position += token.length();
                    break;
                }
            }

            if (!found) {
                throw new RuntimeException("Unexpected sequence found in input string, at " + position);
            }
        }

        return tokens;

    }
}

創建LIKE到正則表達式的轉換映射

public class LikeTranspiler {
    private static Tokenizer TOKENIZER = new Tokenizer()
            .add("^(\\[[^]]*])", ConstantToken::new)
            .add("^(%)", WildcardToken::new)
            .add("^(_)", WildcharToken::new)
            .add("^([^\\[\\]%_]+)", StringToken::new);

    public static String toRegEx(String pattern) throws ParseException {
        StringBuilder sb = new StringBuilder().append("^");
        for (Token token : TOKENIZER.tokenize(pattern)) {
            sb.append(token.convert());
        }

        return sb.append("$").toString();
    }
}

直接調用LikeTranspiler的toRegEx方法將LIKE語法轉爲JAVA中的正則表達式。

private ExprObj calcBinaryOperation(OperationType type, InnerTable table, InnerRow row) {
        ExprObj result = null;
        switch (type) {
            . . .
            case T_OP_LIKE:
                result = logicalCalculus(table, row, (a, b) -> ObjUtil.like(a, b));
                break;
            . . .
        }

        return result;
    }
public static boolean like(Obj first, Obj second) {
        Assert.state(first.getType() == ObjType.STRING, OperationType.T_OP_LIKE + " only support STRING.");
        Assert.state(second.getType() == ObjType.STRING, OperationType.T_OP_LIKE + " only support STRING.");

        String firstValue = (String) first.getRelValue();

        String secondValue = (String) second.getRelValue();

        String regEx = LikeTranspiler.toRegEx(secondValue);

        return Pattern.compile(regEx).matcher(firstValue).matches();
    }

通過創建詞法分析器並使用此方法進行轉換，我們可以防止LIKE像這樣的子句被轉換爲正則表達式%abc[%]%，該子句應將其中的任何子字符串與其中的子字符串匹配，該子句將與子字符串或匹配任何字符串。abc%.abc[.].abc.abc。

類型計算轉換

不同數據類型在進行計算時需要轉型，具體的轉化入下二維數組中。

// 不同類型計算後的類型
ObjType[][] RESULT_TYPE = {
        //UNKNOWN  BYTE     SHORT    INT      LONG     FLOAT    DOUBLE   DECIMAL  BOOL     DATE       TIME       TIMESTAMP  STRING     NULL
        { UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN,   UNKNOWN,   UNKNOWN,   UNKNOWN,   UNKNOWN },// UNKNOWN
        { UNKNOWN, LONG,    LONG,    LONG,    LONG,    DOUBLE,  DOUBLE,  DECIMAL, BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   LONG,      UNKNOWN },// BYTE
        { UNKNOWN, LONG,    LONG,    LONG,    LONG,    DOUBLE,  DOUBLE,  DECIMAL, BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   LONG,      UNKNOWN },// SHORT
        { UNKNOWN, LONG,    LONG,    LONG,    LONG,    DOUBLE,  DOUBLE,  DECIMAL, BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   LONG,      UNKNOWN },// INT
        { UNKNOWN, LONG,    LONG,    LONG,    LONG,    DOUBLE,  DOUBLE,  DECIMAL, BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   LONG,      UNKNOWN },// LONG
        { UNKNOWN, DOUBLE,  DOUBLE,  DOUBLE,  DOUBLE,  DOUBLE,  DOUBLE,  DECIMAL, BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   DOUBLE,    UNKNOWN },// FLOAT
        { UNKNOWN, DOUBLE,  DOUBLE,  DOUBLE,  DOUBLE,  DOUBLE,  DOUBLE,  DECIMAL, BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   DOUBLE,    UNKNOWN },// DOUBLE
        { UNKNOWN, DECIMAL, DECIMAL, DECIMAL, DECIMAL, DECIMAL, DECIMAL, DECIMAL, UNKNOWN, UNKNOWN,   UNKNOWN,   UNKNOWN,   DECIMAL,   UNKNOWN },// DECIMAL
        { UNKNOWN, BOOL,    BOOL,    BOOL,    BOOL,    BOOL,    BOOL,    BOOL,    BOOL,    UNKNOWN,   UNKNOWN,   UNKNOWN,   BOOL,      UNKNOWN },// BOOL
        { UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, TIMESTAMP, TIMESTAMP, TIMESTAMP, TIMESTAMP, UNKNOWN },// DATE
        { UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, TIMESTAMP, TIMESTAMP, TIMESTAMP, TIMESTAMP, UNKNOWN },// TIME
        { UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, TIMESTAMP, TIMESTAMP, TIMESTAMP, TIMESTAMP, UNKNOWN },// TIMESTAMP
        { UNKNOWN, LONG,    LONG,    LONG,    LONG,    DOUBLE,  DOUBLE,  DECIMAL, BOOL,    TIMESTAMP, TIMESTAMP, TIMESTAMP, STRING,    UNKNOWN },// STRING
        { UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN, UNKNOWN,   UNKNOWN,   UNKNOWN,   UNKNOWN,   UNKNOWN },// NULL
};

參考資料

[1] https://codereview.stackexchange.com/questions/36861/convert-sql-like-to-regex/207486

本文分享自華爲雲社區《微服務緩存中間件CoralCache表達式計算引擎詳解》，原文作者：超純的小白兔。

點擊關注，第一時間瞭解華爲雲新鮮技術~

CoralCache：一個提高微服務可用性的中間件

背景

架構

表達式計算引擎

常見運算符的實現

邏輯運算

數學運算

LIKE運算符

LIKE表達式語法

LIKE詞法定義

定義Token類

創建Lexer(Tokenizer)

創建LIKE到正則表達式的轉換映射

類型計算轉換

參考資料

如何使用 JS 判斷用戶是否處於活躍狀態

lightdb秒級增加列和刪除列（not null帶默認值）

lightdb數據庫超時相關控制參數

通過HPA+CronHPA組合應對業務複雜彈性伸縮場景

❤️‍🔥 Solon Cloud Event 新的事務特性與應用

lightdb mysql 8.0兼容之不可見主鍵

使用 JS 實現在瀏覽器控制檯打印圖片 console.image()

基於Ubuntu-22.04安裝K8s-v1.28.2實驗（四）使用域名訪問網站應用

Linux中的tty和pts

網絡爬蟲的祕密：如何高效地抓取JD.com視頻鏈接

系統國際化之多語言解決方案| 京東物流技術團隊

CaffeineCache Api介紹以及與Guava Cache性能對比| 京東物流技術團隊

Vite 的預構建原理與實踐| 京東物流技術團隊

https://yachay.unat.edu.pe/blog/index.php?comment_area=format_blog&comment_component=blog&comment_co

linux以太網驅動總結