org.bson.json.JsonScanner Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of mongo-java-driver Show documentation
Show all versions of mongo-java-driver Show documentation
The MongoDB Java Driver uber-artifact, containing mongodb-driver, mongodb-driver-core, and bson
/*
* Copyright 2008-present MongoDB, Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.bson.json;
import org.bson.BsonRegularExpression;
import java.io.Reader;
/**
* Parses the string representation of a JSON object into a set of {@link JsonToken}-derived objects.
*
* @since 3.0
*/
class JsonScanner {
private final JsonBuffer buffer;
JsonScanner(final JsonBuffer buffer) {
this.buffer = buffer;
}
JsonScanner(final String json) {
this(new JsonStringBuffer(json));
}
JsonScanner(final Reader reader) {
this(new JsonStreamBuffer(reader));
}
public void reset(final int markPos) {
buffer.reset(markPos);
}
public int mark() {
return buffer.mark();
}
public void discard(final int markPos) {
buffer.discard(markPos);
}
/**
* Finds and returns the next complete token from this scanner. If scanner reached the end of the source, it will return a token with
* {@code JSONTokenType.END_OF_FILE} type.
*
* @return The next token.
* @throws JsonParseException if source is invalid.
*/
public JsonToken nextToken() {
int c = buffer.read();
while (c != -1 && Character.isWhitespace(c)) {
c = buffer.read();
}
if (c == -1) {
return new JsonToken(JsonTokenType.END_OF_FILE, "");
}
switch (c) {
case '{':
return new JsonToken(JsonTokenType.BEGIN_OBJECT, "{");
case '}':
return new JsonToken(JsonTokenType.END_OBJECT, "}");
case '[':
return new JsonToken(JsonTokenType.BEGIN_ARRAY, "[");
case ']':
return new JsonToken(JsonTokenType.END_ARRAY, "]");
case '(':
return new JsonToken(JsonTokenType.LEFT_PAREN, "(");
case ')':
return new JsonToken(JsonTokenType.RIGHT_PAREN, ")");
case ':':
return new JsonToken(JsonTokenType.COLON, ":");
case ',':
return new JsonToken(JsonTokenType.COMMA, ",");
case '\'':
case '"':
return scanString((char) c);
case '/':
return scanRegularExpression();
default:
if (c == '-' || Character.isDigit(c)) {
return scanNumber((char) c);
} else if (c == '$' || c == '_' || Character.isLetter(c)) {
return scanUnquotedString((char) c);
} else {
int position = buffer.getPosition();
buffer.unread(c);
throw new JsonParseException("Invalid JSON input. Position: %d. Character: '%c'.", position, c);
}
}
}
/**
* Reads {@code RegularExpressionToken} from source. The following variants of lexemes are possible:
*
* /pattern/
* /\(pattern\)/
* /pattern/ims
*
* Options can include 'i','m','x','s'
*
* @return The regular expression token.
* @throws JsonParseException if regular expression representation is not valid.
*/
private JsonToken scanRegularExpression() {
StringBuilder patternBuilder = new StringBuilder();
StringBuilder optionsBuilder = new StringBuilder();
RegularExpressionState state = RegularExpressionState.IN_PATTERN;
while (true) {
int c = buffer.read();
switch (state) {
case IN_PATTERN:
switch (c) {
case -1:
state = RegularExpressionState.INVALID;
break;
case '/':
state = RegularExpressionState.IN_OPTIONS;
break;
case '\\':
state = RegularExpressionState.IN_ESCAPE_SEQUENCE;
break;
default:
state = RegularExpressionState.IN_PATTERN;
break;
}
break;
case IN_ESCAPE_SEQUENCE:
state = RegularExpressionState.IN_PATTERN;
break;
case IN_OPTIONS:
switch (c) {
case 'i':
case 'm':
case 'x':
case 's':
state = RegularExpressionState.IN_OPTIONS;
break;
case ',':
case '}':
case ']':
case ')':
case -1:
state = RegularExpressionState.DONE;
break;
default:
if (Character.isWhitespace(c)) {
state = RegularExpressionState.DONE;
} else {
state = RegularExpressionState.INVALID;
}
break;
}
break;
default:
break;
}
switch (state) {
case DONE:
buffer.unread(c);
BsonRegularExpression regex
= new BsonRegularExpression(patternBuilder.toString(), optionsBuilder.toString());
return new JsonToken(JsonTokenType.REGULAR_EXPRESSION, regex);
case INVALID:
throw new JsonParseException("Invalid JSON regular expression. Position: %d.", buffer.getPosition());
default:
switch (state) {
case IN_OPTIONS:
if (c != '/') {
optionsBuilder.append((char) c);
}
break;
default:
patternBuilder.append((char) c);
break;
}
}
}
}
/**
* Reads {@code StringToken} from source.
*
* @return The string token.
*/
private JsonToken scanUnquotedString(final char firstChar) {
StringBuilder sb = new StringBuilder();
sb.append(firstChar);
int c = buffer.read();
while (c == '$' || c == '_' || Character.isLetterOrDigit(c)) {
sb.append((char) c);
c = buffer.read();
}
buffer.unread(c);
String lexeme = sb.toString();
return new JsonToken(JsonTokenType.UNQUOTED_STRING, lexeme);
}
/**
* Reads number token from source. The following variants of lexemes are possible:
*
* 12
* 123
* -0
* -345
* -0.0
* 0e1
* 0e-1
* -0e-1
* 1e12
* -Infinity
*
*
* @return The number token.
* @throws JsonParseException if number representation is invalid.
*/
//CHECKSTYLE:OFF
private JsonToken scanNumber(final char firstChar) {
int c = firstChar;
StringBuilder sb = new StringBuilder();
sb.append(firstChar);
NumberState state;
switch (c) {
case '-':
state = NumberState.SAW_LEADING_MINUS;
break;
case '0':
state = NumberState.SAW_LEADING_ZERO;
break;
default:
state = NumberState.SAW_INTEGER_DIGITS;
break;
}
JsonTokenType type = JsonTokenType.INT64;
while (true) {
c = buffer.read();
switch (state) {
case SAW_LEADING_MINUS:
switch (c) {
case '0':
state = NumberState.SAW_LEADING_ZERO;
break;
case 'I':
state = NumberState.SAW_MINUS_I;
break;
default:
if (Character.isDigit(c)) {
state = NumberState.SAW_INTEGER_DIGITS;
} else {
state = NumberState.INVALID;
}
break;
}
break;
case SAW_LEADING_ZERO:
switch (c) {
case '.':
state = NumberState.SAW_DECIMAL_POINT;
break;
case 'e':
case 'E':
state = NumberState.SAW_EXPONENT_LETTER;
break;
case ',':
case '}':
case ']':
case ')':
case -1:
state = NumberState.DONE;
break;
default:
if (Character.isDigit(c)) {
state = NumberState.SAW_INTEGER_DIGITS;
} else if (Character.isWhitespace(c)) {
state = NumberState.DONE;
} else {
state = NumberState.INVALID;
}
break;
}
break;
case SAW_INTEGER_DIGITS:
switch (c) {
case '.':
state = NumberState.SAW_DECIMAL_POINT;
break;
case 'e':
case 'E':
state = NumberState.SAW_EXPONENT_LETTER;
break;
case ',':
case '}':
case ']':
case ')':
case -1:
state = NumberState.DONE;
break;
default:
if (Character.isDigit(c)) {
state = NumberState.SAW_INTEGER_DIGITS;
} else if (Character.isWhitespace(c)) {
state = NumberState.DONE;
} else {
state = NumberState.INVALID;
}
break;
}
break;
case SAW_DECIMAL_POINT:
type = JsonTokenType.DOUBLE;
if (Character.isDigit(c)) {
state = NumberState.SAW_FRACTION_DIGITS;
} else {
state = NumberState.INVALID;
}
break;
case SAW_FRACTION_DIGITS:
switch (c) {
case 'e':
case 'E':
state = NumberState.SAW_EXPONENT_LETTER;
break;
case ',':
case '}':
case ']':
case ')':
case -1:
state = NumberState.DONE;
break;
default:
if (Character.isDigit(c)) {
state = NumberState.SAW_FRACTION_DIGITS;
} else if (Character.isWhitespace(c)) {
state = NumberState.DONE;
} else {
state = NumberState.INVALID;
}
break;
}
break;
case SAW_EXPONENT_LETTER:
type = JsonTokenType.DOUBLE;
switch (c) {
case '+':
case '-':
state = NumberState.SAW_EXPONENT_SIGN;
break;
default:
if (Character.isDigit(c)) {
state = NumberState.SAW_EXPONENT_DIGITS;
} else {
state = NumberState.INVALID;
}
break;
}
break;
case SAW_EXPONENT_SIGN:
if (Character.isDigit(c)) {
state = NumberState.SAW_EXPONENT_DIGITS;
} else {
state = NumberState.INVALID;
}
break;
case SAW_EXPONENT_DIGITS:
switch (c) {
case ',':
case '}':
case ']':
case ')':
state = NumberState.DONE;
break;
default:
if (Character.isDigit(c)) {
state = NumberState.SAW_EXPONENT_DIGITS;
} else if (Character.isWhitespace(c)) {
state = NumberState.DONE;
} else {
state = NumberState.INVALID;
}
break;
}
break;
case SAW_MINUS_I:
boolean sawMinusInfinity = true;
char[] nfinity = new char[]{'n', 'f', 'i', 'n', 'i', 't', 'y'};
for (int i = 0; i < nfinity.length; i++) {
if (c != nfinity[i]) {
sawMinusInfinity = false;
break;
}
sb.append((char) c);
c = buffer.read();
}
if (sawMinusInfinity) {
type = JsonTokenType.DOUBLE;
switch (c) {
case ',':
case '}':
case ']':
case ')':
case -1:
state = NumberState.DONE;
break;
default:
if (Character.isWhitespace(c)) {
state = NumberState.DONE;
} else {
state = NumberState.INVALID;
}
break;
}
} else {
state = NumberState.INVALID;
}
break;
default:
}
switch (state) {
case INVALID:
throw new JsonParseException("Invalid JSON number");
case DONE:
buffer.unread(c);
String lexeme = sb.toString();
if (type == JsonTokenType.DOUBLE) {
return new JsonToken(JsonTokenType.DOUBLE, Double.parseDouble(lexeme));
} else {
long value = Long.parseLong(lexeme);
if (value < Integer.MIN_VALUE || value > Integer.MAX_VALUE) {
return new JsonToken(JsonTokenType.INT64, value);
} else {
return new JsonToken(JsonTokenType.INT32, (int) value);
}
}
default:
sb.append((char) c);
}
}
}
//CHECKSTYLE:ON
/**
* Reads {@code StringToken} from source.
*
* @return The string token.
*/
//CHECKSTYLE:OFF
private JsonToken scanString(final char quoteCharacter) {
StringBuilder sb = new StringBuilder();
while (true) {
int c = buffer.read();
switch (c) {
case '\\':
c = buffer.read();
switch (c) {
case '\'':
sb.append('\'');
break;
case '"':
sb.append('"');
break;
case '\\':
sb.append('\\');
break;
case '/':
sb.append('/');
break;
case 'b':
sb.append('\b');
break;
case 'f':
sb.append('\f');
break;
case 'n':
sb.append('\n');
break;
case 'r':
sb.append('\r');
break;
case 't':
sb.append('\t');
break;
case 'u':
int u1 = buffer.read();
int u2 = buffer.read();
int u3 = buffer.read();
int u4 = buffer.read();
if (u4 != -1) {
String hex = new String(new char[]{(char) u1, (char) u2, (char) u3, (char) u4});
sb.append((char) Integer.parseInt(hex, 16));
}
break;
default:
throw new JsonParseException("Invalid escape sequence in JSON string '\\%c'.", c);
}
break;
default:
if (c == quoteCharacter) {
return new JsonToken(JsonTokenType.STRING, sb.toString());
}
if (c != -1) {
sb.append((char) c);
}
}
if (c == -1) {
throw new JsonParseException("End of file in JSON string.");
}
}
}
private enum NumberState {
SAW_LEADING_MINUS,
SAW_LEADING_ZERO,
SAW_INTEGER_DIGITS,
SAW_DECIMAL_POINT,
SAW_FRACTION_DIGITS,
SAW_EXPONENT_LETTER,
SAW_EXPONENT_SIGN,
SAW_EXPONENT_DIGITS,
SAW_MINUS_I,
DONE,
INVALID
}
private enum RegularExpressionState {
IN_PATTERN,
IN_ESCAPE_SEQUENCE,
IN_OPTIONS,
DONE,
INVALID
}
}
© 2015 - 2024 Weber Informatics LLC | Privacy Policy