1 /*
2     This file is part of the KDE libraries
3 
4     SPDX-FileCopyrightText: 2003, 2007 Oswald Buddenhagen <ossi@kde.org>
5 
6     SPDX-License-Identifier: LGPL-2.0-or-later
7 */
8 #ifndef KSHELL_H
9 #define KSHELL_H
10 
11 #include <QStringList>
12 #include <kcoreaddons_export.h>
13 #include <qglobal.h>
14 
15 class QString;
16 
17 /**
18  * \namespace KShell
19  * Emulates some basic system shell functionality.
20  * @see KStringHandler
21  */
22 namespace KShell
23 {
24 /**
25  * Flags for splitArgs().
26  * @see Options
27  */
28 enum Option {
29     NoOptions = 0,
30 
31     /**
32      * Perform tilde expansion.
33      * On Windows, this flag is ignored, as the Windows shell has no
34      * equivalent functionality.
35      */
36     TildeExpand = 1,
37 
38     /**
39      * Put the parser into full shell mode and bail out if a too complex
40      * construct is encountered.
41      * A particular purpose of this flag is finding out whether the
42      * command line being split would be executable directly (via
43      * KProcess::setProgram()) or whether it needs to be run through
44      * a real shell (via KProcess::setShellCommand()). Note, however,
45      * that shell builtins are @em not recognized - you need to do that
46      * yourself (compare with a list of known commands or verify that an
47      * executable exists for the named command).
48      *
49      * Meta characters that cause a bail-out are the command separators
50      * @c semicolon and @c ampersand, the redirection symbols @c less-than,
51      * @c greater-than and the @c pipe @c symbol and the grouping symbols
52      * opening and closing @c parentheses.
53      *
54      * Further meta characters on *NIX are the grouping symbols
55      * opening and closing @c braces, the command substitution symbol
56      * @c backquote, the generic substitution symbol @c dollar (if
57      * not followed by an apostrophe), the wildcards @c asterisk,
58      * @c question @c mark and opening and closing @c square @c brackets
59      * and the comment symbol @c hash @c mark.
60      * Additionally, a variable assignment in the first word is recognized.
61      *
62      * A further meta character on Windows is the environment variable
63      * expansion symbol @c percent. Occurrences of @c \%PERCENT_SIGN% as
64      * inserted by quoteArg() are converted back and cause no bail-out,
65      * though.
66      */
67     AbortOnMeta = 2,
68 };
69 /**
70  * Stores a combination of #Option values.
71  */
72 Q_DECLARE_FLAGS(Options, Option)
73 Q_DECLARE_OPERATORS_FOR_FLAGS(Options)
74 
75 /**
76  * Status codes from splitArgs()
77  */
78 enum Errors {
79     /**
80      * Success.
81      */
82     NoError = 0,
83 
84     /**
85      * Indicates a parsing error, like an unterminated quoted string.
86      */
87     BadQuoting,
88 
89     /**
90      * The AbortOnMeta flag was set and an unhandled shell meta character
91      * was encountered.
92      */
93     FoundMeta,
94 };
95 
96 /**
97  * Splits @p cmd according to system shell word splitting and quoting rules.
98  * Can optionally perform tilde expansion and/or abort if it finds shell
99  * meta characters it cannot process.
100  *
101  * On *NIX the behavior is based on the POSIX shell and bash:
102  * - Whitespace splits tokens
103  * - The backslash quotes the following character
104  * - A string enclosed in single quotes is not split. No shell meta
105  *   characters are interpreted.
106  * - A string enclosed in double quotes is not split. Within the string,
107  *   the backslash quotes shell meta characters - if it is followed
108  *   by a "meaningless" character, the backslash is output verbatim.
109  * - A string enclosed in $'' is not split. Within the string, the
110  *   backslash has a similar meaning to the one in C strings. Consult
111  *   the bash manual for more information.
112  *
113  * On Windows, the behavior is defined by the Microsoft C runtime. Qt and
114  * many other implementations comply with this standard, but many do not.
115  * - Whitespace splits tokens
116  * - A string enclosed in double quotes is not split
117  *   - 2N double quotes within a quoted string yield N literal quotes.
118  *     This is not documented on MSDN.
119  * - Backslashes have special semantics iff they are followed by a double
120  *   quote:
121  *   - 2N backslashes + double quote => N backslashes and begin/end quoting
122  *   - 2N+1 backslashes + double quote => N backslashes + literal quote
123  *
124  * If AbortOnMeta is used on Windows, this function applies cmd shell
125  * semantics before proceeding with word splitting:
126  * - Cmd ignores @em all special chars between double quotes.
127  *   Note that the quotes are @em not removed at this stage - the
128  *   tokenization rules described above still apply.
129  * - The @c circumflex is the escape char for everything including
130  *   itself.
131  *
132  * @param cmd the command to split
133  * @param flags operation flags, see \ref Option
134  * @param err if not NULL, a status code will be stored at the pointer
135  *  target, see \ref Errors
136  * @return a list of unquoted words or an empty list if an error occurred
137  */
138 KCOREADDONS_EXPORT QStringList splitArgs(const QString &cmd, Options flags = NoOptions, Errors *err = nullptr);
139 
140 /**
141  * Quotes and joins @p args together according to system shell rules.
142  *
143  * If the output is fed back into splitArgs(), the AbortOnMeta flag
144  * needs to be used on Windows. On *NIX, no such requirement exists.
145  *
146  * See quoteArg() for more info.
147  *
148  * @param args a list of strings to quote and join
149  * @return a command suitable for shell execution
150  */
151 KCOREADDONS_EXPORT QString joinArgs(const QStringList &args);
152 
153 /**
154  * Quotes @p arg according to system shell rules.
155  *
156  * This function can be used to quote an argument string such that
157  * the shell processes it properly. This is e.g. necessary for
158  * user-provided file names which may contain spaces or quotes.
159  * It also prevents expansion of wild cards and environment variables.
160  *
161  * On *NIX, the output is POSIX shell compliant.
162  * On Windows, it is compliant with the argument splitting code of the
163  * Microsoft C runtime and the cmd shell used together.
164  * Occurrences of the @c percent @c sign are replaced with
165  * @c \%PERCENT_SIGN% to prevent spurious variable expansion;
166  * related KDE functions are prepared for this.
167  *
168  * @param arg the argument to quote
169  * @return the quoted argument
170  */
171 KCOREADDONS_EXPORT QString quoteArg(const QString &arg);
172 
173 /**
174  * Performs tilde expansion on @p path. Interprets "~/path" and
175  * "~user/path". If the path starts with an escaped tilde ("\~" on UNIX,
176  * "^~" on Windows), the escape char is removed and the path is returned
177  * as is.
178  *
179  * Note that if @p path starts with a tilde but cannot be properly expanded,
180  * this function will return an empty string.
181  *
182  * @param path the path to tilde-expand
183  * @return the expanded path
184  */
185 KCOREADDONS_EXPORT QString tildeExpand(const QString &path);
186 
187 /**
188  * Performs tilde collapse on @p path. If path did not start by the user
189  * homedir returns path unchanged.
190  *
191  * @param path the path to tilde-collpase
192  * @return the collapsed path
193  * @since 5.67
194  */
195 KCOREADDONS_EXPORT QString tildeCollapse(const QString &path);
196 }
197 
198 #endif /* KSHELL_H */
199