/* -*- mode: java; c-basic-offset: 2; indent-tabs-mode: nil -*- */

/*

Part of the Processing project - http://processing.org

Original Copyright (c) 1997, 1998 Van Di-Han HO. All Rights Reserved.

Updates Copyright (c) 2001 Jason Pell.

Further updates Copyright (c) 2003 Martin Gomez, Ateneo de Manila University

Additional updates Copyright (c) 2005-10 Ben Fry and Casey Reas

Even more updates Copyright (c) 2014 George Bateman

This program is free software; you can redistribute it and/or modify

it under the terms of the GNU General Public License as published by

the Free Software Foundation, version 2.

This program is distributed in the hope that it will be useful,

but WITHOUT ANY WARRANTY; without even the implied warranty of

MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

GNU General Public License for more details.

You should have received a copy of the GNU General Public License

along with this program; if not, write to the Free Software Foundation,

Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA

*/

package processing.mode.java;

import java.util.Stack;

import java.util.regex.Pattern;

import processing.app.Formatter;

import processing.app.Preferences;

import processing.core.PApplet;

/**

* Handler for dealing with auto format.

* Contributed by Martin Gomez, additional bug fixes by Ben Fry.

* Additional fixes by Jonathan Feinberg in March 2010.

* <p/>

* After some further digging, this code in fact appears to be a modified

* version of Jason Pell's GPLed "Java Beautifier" class found

* <a href="http://web.archive.org/web/20091018220353/http://geocities.com/jasonpell/programs.html">here</a>,

* which is itself based on code from Van Di-Han Ho from

* <a href="http://web.archive.org/web/20091027043013/http://www.geocities.com/~starkville/vancbj_idx.html">here</a>.

* [Ben Fry, August 2009]

*/

public class AutoFormat implements Formatter {

private char[] chars;

private final StringBuilder buf = new StringBuilder();

private final StringBuilder result = new StringBuilder();

/** The number of spaces in one indent. Constant. */

private int indentValue;

/** Set when the end of the chars array is reached. */

private boolean EOF;

private boolean inStatementFlag; // in a line of code

private boolean overflowFlag; // line overrunning?

private boolean startFlag; // No buf has yet been writen to this line.

private boolean if_flg;

private boolean elseFlag;

/** -1 if not in array or if just after it, otherwise increases from 0. */

private int arrayLevel;

private int arrayIndent; // Lowest value of the above for this line.

/** Number of ? entered without exiting at : of a?b:c structures. */

private int conditionalLevel;

private int[][] sp_flg;

private boolean[][] s_ind;

private int if_lev;

/** chars[pos] is where we're at. */

private int pos;

private int level;

/** Number of curly brackets entered and not exited,

excluding arrays. */

private int curlyLvl;

/** Number of parentheses entered and not exited. */

private int parenLevel;

private boolean[] ind;

private int[] p_flg;

private int[][] s_tabs;

/** At every {, this has a true pushed if it's a do-while,

and a false otherwise. It is then popped at }. */

private Stack<Boolean> doWhileFlags;

/** At every (, this has a true pushed for if, while, or for,

and a false otherwise. Popped at ). */

private Stack<Boolean> ifWhileForFlags;

private boolean jdoc_flag;

/** The number of times to indent at a given point */

private int tabs;

/** The last non-space seen by nextChar(). */

private char lastNonWhitespace = 0;

private void handleMultiLineComment() {

final boolean savedStartFlag = startFlag;

buf.append(nextChar()); // So /*/ isn't self-closing.

for (char ch = nextChar(); !EOF; ch = nextChar()) {

buf.append(ch);

while (ch != '/' && !EOF) {

if (ch == '\n') {

writeIndentedComment();

startFlag = true;

}

buf.append(ch = nextChar());

}

if (buf.length() >= 2 && buf.charAt(buf.length() - 2) == '*') {

jdoc_flag = false;

break;

}

}

writeIndentedComment();

startFlag = savedStartFlag;

jdoc_flag = false;

return;

}

/**

* Pumps nextChar into buf until \n or EOF, then calls

* writeIndentedLine() and sets startFlag to true.

*/

private void handleSingleLineComment() {

char ch = nextChar();

while (ch != '\n' && !EOF) {

buf.append(ch);

ch = nextChar();

}

writeIndentedLine();

startFlag = true;

}

private void writeIndentedLine() {

if (buf.length() == 0) {

if (startFlag) startFlag = elseFlag = false;

return;

}

if (startFlag) {

// Indent suppressed at eg. if<nl>{ and when

// buf is close-brackets only followed by ';'.

boolean indentMore = !buf.toString().matches("[\\s\\]\\}\\)]+;")

&& (buf.charAt(0) != '{' || arrayLevel >= 0)

&& overflowFlag;

if (indentMore) {

tabs++;

if (arrayIndent > 0) tabs += arrayIndent;

}

printIndentation();

startFlag = false;

if (indentMore) {

tabs--;

if (arrayIndent > 0) tabs -= arrayIndent;

}

}

if (lastNonSpaceChar() == '}' && bufStarts("else")) {

result.append(' ');

}

if (elseFlag) {

if (lastNonSpaceChar() == '}') {

trimRight(result);

result.append(' ');

}

elseFlag = false;

}

// If we're still in a statement at \n, that's overflow.

overflowFlag = inStatementFlag;

arrayIndent = arrayLevel;

result.append(buf);

buf.setLength(0);

}

/**

* @return the last character in <tt>result</tt> not ' ' or '\n'.

*/

private char lastNonSpaceChar() {

for (int i = result.length() - 1; i >= 0; i--) {

char chI = result.charAt(i);

if (chI != ' ' && chI != '\n') return chI;

}

return 0;

}

/**

* Called by handleMultilineComment.<br />

* Sets jdoc_flag if at the start of a doc comment.

* Sends buf to result with proper indents, then clears buf.<br />

* Does nothing if buf is empty.

*/

private void writeIndentedComment() {

if (buf.length() == 0) return;

int firstNonSpace = 0;

while (buf.charAt(firstNonSpace) == ' ') firstNonSpace++;

if (lookup_com("/**")) jdoc_flag = true;

if (startFlag) printIndentation();

if (buf.charAt(firstNonSpace) == '/' && buf.charAt(firstNonSpace+1) == '*') {

if (startFlag && lastNonWhitespace != ';') {

result.append(buf.substring(firstNonSpace));

} else {

result.append(buf);

}

} else {

if (buf.charAt(firstNonSpace) == '*' || !jdoc_flag) {

result.append(" " + buf.substring(firstNonSpace));

} else {

result.append(" * " + buf.substring(firstNonSpace));

}

}

buf.setLength(0);

}

/**

* Makes tabs &gt;= 0 and appends <tt>tabs*indentValue</tt>

* spaces to result.

*/

private void printIndentation() {

if (tabs <= 0) {

tabs = 0;

return;

}

final int spaces = tabs * indentValue;

for (int i = 0; i < spaces; i++) {

result.append(' ');

}

}

/**

* @return <tt>chars[pos+1]</tt> or '\0' if out-of-bounds.

*/

private char peek() {

return (pos + 1 >= chars.length) ? 0 : chars[pos + 1];

}

/**

* Sets pos to the position of the next character that is not ' '

* in chars. If chars[pos] != ' ' already, it will still move on.

* Then sets EOF if pos has reached the end, or reverses pos by 1 if it

* has not.

* <br/> Does nothing if EOF.

* @param allWsp (boolean) Eat newlines too (all of Character.isWhiteSpace()).

*/

private void advanceToNonSpace(boolean allWsp) {

if (EOF) return;

if (allWsp) {

do {

pos++;

} while (pos < chars.length && Character.isWhitespace(chars[pos]));

} else {

do {

pos++;

} while (pos < chars.length && chars[pos] == ' ');

}

if (pos == chars.length - 1) {

EOF = true;

} else {

pos--; // reset for nextChar()

}

}

/**

* Increments pos, sets EOF if needed, and returns the new

* chars[pos] or zero if out-of-bounds.

* Sets lastNonWhitespace if chars[pos] isn't whitespace.

* Does nothing and returns zero if already at EOF.

*/

private char nextChar() {

if (EOF) return 0;

pos++;

if (pos >= chars.length - 1) EOF = true;

if (pos >= chars.length) return 0;

char retVal = chars[pos];

if (!Character.isWhitespace(retVal)) lastNonWhitespace = retVal;

return retVal;

}

/**

* Called after else.

*/

private void gotElse() {

tabs = s_tabs[curlyLvl][if_lev];

p_flg[level] = sp_flg[curlyLvl][if_lev];

ind[level] = s_ind[curlyLvl][if_lev];

if_flg = true;

// We can't expect else to be followed by a semicolon, so must

// end the statemant manually.

inStatementFlag = false;

}

/**

* Pump any '\t' and ' ' to buf, handle any following comment,

* and if the next character is '\n', discard it.

* @return Whether a '\n' was found and discarded.

*/

private boolean readForNewLine() {

final int savedTabs = tabs;

char c = peek();

while (!EOF && (c == '\t' || c == ' ')) {

buf.append(nextChar());

c = peek();

}

if (c == '/') {

buf.append(nextChar());

c = peek();

if (c == '*') {

buf.append(nextChar());

handleMultiLineComment();

} else if (c == '/') {

buf.append(nextChar());

handleSingleLineComment();

return true;

}

}

c = peek();

if (c == '\n') {

// eat it

nextChar();

tabs = savedTabs;

return true;

}

return false;

}

/**

* @return last non-wsp in result+buf, or 0 on error.

*/

private char prevNonWhitespace() {

StringBuffer tot = new StringBuffer();

tot.append(result);

tot.append(buf);

for (int i = tot.length()-1; i >= 0; i--) {

if (!Character.isWhitespace(tot.charAt(i)))

return tot.charAt(i);

}

return 0;

}

/**

* Sees if buf is of the form [optional whitespace][keyword][optional anything].

* It won't allow keyword to be directly followed by an alphanumeric, _, or &amp;.

* Will be different if keyword contains regex codes.

*/

private boolean bufStarts(final String keyword) {

return Pattern.matches("^\\s*" + keyword + "(?![a-zA-Z0-9_&]).*$", buf);

}

/**

* Sees if buf is of the form [optional anything][keyword][optional whitespace].

* It won't allow keyword to be directly preceded by an alphanumeric, _, or &amp;.

* Will be different if keyword contains regex codes.

*/

private boolean bufEnds(final String keyword) {

return Pattern.matches("^.*(?<![a-zA-Z0-9_&])" + keyword + "\\s*$", buf);

}

/**

* Allows you to increase if_lev safely. Enlarges related arrays where needed;

* does not change if_lev itself.

*/

private void if_levSafe() {

if (s_tabs[0].length <= if_lev) {

for (int i = 0; i < s_tabs.length; i++) {

s_tabs[i] = PApplet.expand(s_tabs[i]);

}

}

if (sp_flg[0].length <= if_lev) {

for (int i = 0; i < sp_flg.length; i++) {

sp_flg[i] = PApplet.expand(sp_flg[i]);

}

}

if (s_ind[0].length <= if_lev) {

for (int i = 0; i < s_ind.length; i++) {

s_ind[i] = PApplet.expand(s_ind[i]);

}

}

}

/**

* Sees if buf is of the form [optional whitespace][keyword][optional anything].

* It *will* allow keyword to be directly followed by an alphanumeric, _, or &amp;.

* Will be different if keyword contains regex codes (except *, which is fine).

*/

private boolean lookup_com(final String keyword) {

final String regex = "^\\s*" + keyword.replace("*", "\\*") + ".*$";

return Pattern.matches(regex, buf);

}

/**

* Takes all whitespace off the end of its argument.

*/

static private void trimRight(final StringBuilder sb) {

while (sb.length() >= 1 && Character.isWhitespace(sb.charAt(sb.length() - 1))) {

sb.setLength(sb.length() - 1);

}

}

/** Entry point */

public String format(final String source) {

final String normalizedText = source.replaceAll("\r", "");

String cleanText = normalizedText;

if (!normalizedText.endsWith("\n")) {

cleanText += "\n";

}

// Globals' description at top of file.

result.setLength(0);

indentValue = Preferences.getInteger("editor.tabs.size");

boolean forFlag = if_flg = false;

startFlag = true;

int forParenthLevel = 0;

conditionalLevel = parenLevel = curlyLvl = if_lev = level = 0;

tabs = 0;

jdoc_flag = inStatementFlag = overflowFlag = false;

pos = arrayLevel = -1;

int[] s_level = new int[10];

sp_flg = new int[20][10];

s_ind = new boolean[20][10];

int[] s_if_lev = new int[10]; // Stack

boolean[] s_if_flg = new boolean[10]; // Stack

ind = new boolean[10];

p_flg = new int[10];

s_tabs = new int[20][10];

doWhileFlags = new Stack<>();

ifWhileForFlags = new Stack<>();

chars = cleanText.toCharArray();

EOF = false; // set in nextChar() when EOF

while (!EOF) {

char c = nextChar();

switch (c) {

default:

inStatementFlag = true;

buf.append(c);

break;

case ',':

inStatementFlag = true;

trimRight(buf);

buf.append(", ");

advanceToNonSpace(false);

break;

case ' ':

case '\t':

elseFlag = bufEnds("else");

if (elseFlag) {

gotElse();

if (!startFlag || buf.length() > 0) {

buf.append(c);

}

writeIndentedLine();

startFlag = false;

break;

}

// Only allow in the line, to nuke the old indent.

if (!startFlag || buf.length() > 0) buf.append(c);

break;

case '\n':

if (EOF) break;

elseFlag = bufEnds("else");

if (elseFlag) gotElse();

if (lookup_com("//")) {

if (buf.charAt(buf.length() - 1) == '\n') {

buf.setLength(buf.length() - 1);

}

}

if (elseFlag) {

writeIndentedLine();

result.append("\n");

p_flg[level]++;

tabs++;

} else {

writeIndentedLine();

result.append("\n");

}

startFlag = true;

break;

case '{':

elseFlag = bufEnds("else");

if (elseFlag) gotElse();

doWhileFlags.push(Boolean.valueOf(bufEnds("do")));

char prevChar = prevNonWhitespace();

if (arrayLevel >= 0 || prevChar == '=' || prevChar == ']') {

// If we're already in an array (lvl >= 0), increment level.

// Otherwise, the presence of a = or ] indicates an array is starting

// and we should start counting (set lvl=0).

arrayLevel++;

buf.append(c);

break; // Nothing fancy.

}

inStatementFlag = false; // eg. class declaration ends

if (s_if_lev.length == curlyLvl) {

s_if_lev = PApplet.expand(s_if_lev);

s_if_flg = PApplet.expand(s_if_flg);

}

s_if_lev[curlyLvl] = if_lev;

s_if_flg[curlyLvl] = if_flg;

if_lev = 0;

if_flg = false;

curlyLvl++;

if (startFlag && p_flg[level] != 0) {

p_flg[level]--;

tabs--;

}

trimRight(buf);

if (buf.length() > 0 || (result.length() > 0 &&

!Character.isWhitespace(result.charAt(result.length() - 1)))) {

buf.append(" ");

}

buf.append(c);

writeIndentedLine();

readForNewLine();

writeIndentedLine();

result.append('\n');

tabs++;

startFlag = true;

if (p_flg[level] > 0) {

ind[level] = true;

level++;

s_level[level] = curlyLvl;

}

break;

case '}':

if (arrayLevel >= 0) {

// Even less fancy. Note that }s cannot end array behaviour;

// a semicolon is needed.

if (arrayLevel > 0) arrayLevel--;

if (arrayIndent > arrayLevel) arrayIndent = arrayLevel;

buf.append(c);

break;

}

// In a simple enum, there's not necessarily a `;` to end the statement.

inStatementFlag = false;

curlyLvl--;

if (curlyLvl < 0) {

curlyLvl = 0;

buf.append(c);

writeIndentedLine();

} else {

if_lev = s_if_lev[curlyLvl] - 1;

if (if_lev < 0) if_lev = 0;

if_levSafe();

if_flg = s_if_flg[curlyLvl];

trimRight(buf);

writeIndentedLine();

tabs--;

trimRight(result);

result.append('\n');

overflowFlag = false; // Would normally be done in writeIndentedLine.

printIndentation();

result.append(c);

if (peek() == ';') result.append(nextChar());

// doWhileFlags contains a TRUE if and only if the

// corresponding { was preceeded (bufEnds) by "do".

// Just checking for while would fail on if(){} while(){}.

if (doWhileFlags.empty() || !doWhileFlags.pop().booleanValue()

|| !new String(chars, pos+1, chars.length-pos-1).trim().startsWith("while")) {

readForNewLine();

writeIndentedLine();

result.append('\n');

startFlag = true;

} else {

// Correct spacing in "} while".

result.append(' ');

advanceToNonSpace(true);

startFlag = false;

}

if (curlyLvl < s_level[level] && level > 0) level--;

if (ind[level]) {

tabs -= p_flg[level];

p_flg[level] = 0;

ind[level] = false;

}

}

break;

case '"':

case '“':

case '”':

case '\'':

case '‘':

case '’':

inStatementFlag = true;

char realQuote = c;

if (c == '“' || c == '”') realQuote = '"';

if (c == '‘' || c == '’') realQuote = '\'';

buf.append(realQuote);

char otherQuote = c;

if (c == '“') otherQuote = '”';

if (c == '”') otherQuote = '“';

if (c == '‘') otherQuote = '’';

if (c == '’') otherQuote = '‘';

char cc = nextChar();

// In a proper string, all the quotes tested are c. In a curly-quoted

// string, there are three possible end quotes: c, its reverse, and

// the correct straight quote.

while (!EOF && cc != otherQuote && cc != realQuote && cc != c) {

buf.append(cc);

if (cc == '\\') {

buf.append(cc = nextChar());

}

// Syntax error: unterminated string. Leave \n in nextChar, so it

// feeds back into the loop.

if (peek() == '\n') break;

cc = nextChar();

}

if (cc == otherQuote || cc == realQuote || cc == c) {

buf.append(realQuote);

if (readForNewLine()) {

// push a newline into the stream

chars[pos--] = '\n';

}

} else {

// We've had a syntax error if the string wasn't terminated by EOL/

// EOF, just abandon this statement.

inStatementFlag = false;

}

break;

case ';':

if (forFlag) {

// This is like a comma.

trimRight(buf);

buf.append("; ");

// Not non-whitespace: allow \n.

advanceToNonSpace(false);

break;

}

buf.append(c);

inStatementFlag = false;

writeIndentedLine();

if (p_flg[level] > 0 && !ind[level]) {

tabs -= p_flg[level];

p_flg[level] = 0;

}

readForNewLine();

writeIndentedLine();

result.append("\n");

startFlag = true;

// Array behaviour ends at the end of a statement.

arrayLevel = -1;

if (if_lev > 0) {

if (if_flg) {

if_lev--;

if_flg = false;

} else {

if_lev = 0;

}

}

break;

case '\\':

buf.append(c);

buf.append(nextChar());

break;

case '?':

conditionalLevel++;

buf.append(c);

break;

case ':':

// Java 8 :: operator.

if (peek() == ':') {

buf.append(c).append(nextChar());

break;

}

// End a?b:c structures.

else if (conditionalLevel > 0) {

conditionalLevel--;

buf.append(c);

break;

}

else if (forFlag) {

trimRight(buf);

buf.append(" : ");

// Not to non-whitespace: allow \n.

advanceToNonSpace(false);

break;

}

buf.append(c);

inStatementFlag = false;

arrayLevel = -1; // Unlikely to be needed; just in case.

// Same format for case, default, and other labels.

if (tabs > 0) {

tabs--;

writeIndentedLine();

tabs++;

} else {

writeIndentedLine();

}

readForNewLine();

writeIndentedLine();

result.append('\n');

startFlag = true;

break;

case '/':

final char next = peek();

if (next == '/') {

// call nextChar to move on.

buf.append(c).append(nextChar());

handleSingleLineComment();

result.append("\n");

} else if (next == '*') {

if (buf.length() > 0) {

writeIndentedLine();

}

buf.append(c).append(nextChar());

handleMultiLineComment();

} else {

buf.append(c);

}

break;

case ')':

parenLevel--;

// If we're further back than the start of a for loop, we've

// left it.

if (forFlag && forParenthLevel > parenLevel) forFlag = false;

if (parenLevel < 0) parenLevel = 0;

buf.append(c);

boolean wasIfEtc = !ifWhileForFlags.empty() && ifWhileForFlags.pop().booleanValue();

if (wasIfEtc) {

inStatementFlag = false;

arrayLevel = -1; // This is important as it allows arrays in if statements.

}

writeIndentedLine();

// Short-circuiting means readForNewLine is only called for if/while/for;

// this is important.

if (wasIfEtc && readForNewLine()) {

chars[pos] = '\n';

pos--; // Make nextChar() return the new '\n'.

if (parenLevel == 0) {

p_flg[level]++;

tabs++;

ind[level] = false;

}

}

break;

case '(':

final boolean isFor = bufEnds("for");

final boolean isIf = bufEnds("if");

if (isFor || isIf || bufEnds("while")) {

if (!Character.isWhitespace(buf.charAt(buf.length() - 1))) {

buf.append(' ');

}

ifWhileForFlags.push(true);

} else {

ifWhileForFlags.push(false);

}

buf.append(c);

parenLevel++;

// isFor says "Is it the start of a for?". If it is, we set forFlag and

// forParenthLevel. If it is not parenth_lvl was incremented above and

// that's it.

if (isFor && !forFlag) {

forParenthLevel = parenLevel;

forFlag = true;

} else if (isIf) {

writeIndentedLine();

s_tabs[curlyLvl][if_lev] = tabs;

sp_flg[curlyLvl][if_lev] = p_flg[level];

s_ind[curlyLvl][if_lev] = ind[level];

if_lev++;

if_levSafe();

if_flg = true;

}

} // end switch

} // end while not EOF

if (buf.length() > 0) writeIndentedLine();

final String formatted = simpleRegexCleanup(result.toString());

return formatted.equals(cleanText) ? source : formatted;

}

/**

* Make minor regex-based find / replace changes to execute simple fixes to limited artifacts.

*

* @param result The code to format.

* @return The formatted code.

*/

static private String simpleRegexCleanup(String result) {

return result.replaceAll("([^ \n]+) +\n", "$1\n"); // Remove trail whitespace

}

}