TLA Line data Source code
1 : ///
2 : /// Module for parsing postgresql.conf file.
3 : ///
4 : /// NOTE: This doesn't implement the full, correct postgresql.conf syntax. Just
5 : /// enough to extract a few settings we need in Neon, assuming you don't do
6 : /// funny stuff like include-directives or funny escaping.
7 : use anyhow::{bail, Context, Result};
8 : use once_cell::sync::Lazy;
9 : use regex::Regex;
10 : use std::collections::HashMap;
11 : use std::fmt;
12 : use std::io::BufRead;
13 : use std::str::FromStr;
14 :
15 : /// In-memory representation of a postgresql.conf file
16 CBC 568 : #[derive(Default, Debug)]
17 : pub struct PostgresConf {
18 : lines: Vec<String>,
19 : hash: HashMap<String, String>,
20 : }
21 :
22 UBC 0 : static CONF_LINE_RE: Lazy<Regex> = Lazy::new(|| Regex::new(r"^((?:\w|\.)+)\s*=\s*(\S+)$").unwrap());
23 :
24 : impl PostgresConf {
25 CBC 568 : pub fn new() -> PostgresConf {
26 568 : PostgresConf::default()
27 568 : }
28 :
29 : /// Read file into memory
30 UBC 0 : pub fn read(read: impl std::io::Read) -> Result<PostgresConf> {
31 0 : let mut result = Self::new();
32 :
33 0 : for line in std::io::BufReader::new(read).lines() {
34 0 : let line = line?;
35 :
36 : // Store each line in a vector, in original format
37 0 : result.lines.push(line.clone());
38 0 :
39 0 : // Also parse each line and insert key=value lines into a hash map.
40 0 : //
41 0 : // FIXME: This doesn't match exactly the flex/bison grammar in PostgreSQL.
42 0 : // But it's close enough for our usage.
43 0 : let line = line.trim();
44 0 : if line.starts_with('#') {
45 : // comment, ignore
46 0 : continue;
47 0 : } else if let Some(caps) = CONF_LINE_RE.captures(line) {
48 0 : let name = caps.get(1).unwrap().as_str();
49 0 : let raw_val = caps.get(2).unwrap().as_str();
50 :
51 0 : if let Ok(val) = deescape_str(raw_val) {
52 0 : // Note: if there's already an entry in the hash map for
53 0 : // this key, this will replace it. That's the behavior what
54 0 : // we want; when PostgreSQL reads the file, each line
55 0 : // overrides any previous value for the same setting.
56 0 : result.hash.insert(name.to_string(), val.to_string());
57 0 : }
58 0 : }
59 : }
60 0 : Ok(result)
61 0 : }
62 :
63 : /// Return the current value of 'option'
64 0 : pub fn get(&self, option: &str) -> Option<&str> {
65 0 : self.hash.get(option).map(|x| x.as_ref())
66 0 : }
67 :
68 : /// Return the current value of a field, parsed to the right datatype.
69 : ///
70 : /// This calls the FromStr::parse() function on the value of the field. If
71 : /// the field does not exist, or parsing fails, returns an error.
72 : ///
73 : pub fn parse_field<T>(&self, field_name: &str, context: &str) -> Result<T>
74 : where
75 : T: FromStr,
76 : <T as FromStr>::Err: std::error::Error + Send + Sync + 'static,
77 : {
78 0 : self.get(field_name)
79 0 : .with_context(|| format!("could not find '{}' option {}", field_name, context))?
80 0 : .parse::<T>()
81 0 : .with_context(|| format!("could not parse '{}' option {}", field_name, context))
82 0 : }
83 :
84 : pub fn parse_field_optional<T>(&self, field_name: &str, context: &str) -> Result<Option<T>>
85 : where
86 : T: FromStr,
87 : <T as FromStr>::Err: std::error::Error + Send + Sync + 'static,
88 : {
89 0 : if let Some(val) = self.get(field_name) {
90 0 : let result = val
91 0 : .parse::<T>()
92 0 : .with_context(|| format!("could not parse '{}' option {}", field_name, context))?;
93 :
94 0 : Ok(Some(result))
95 : } else {
96 0 : Ok(None)
97 : }
98 0 : }
99 :
100 : ///
101 : /// Note: if you call this multiple times for the same option, the config
102 : /// file will a line for each call. It would be nice to have a function
103 : /// to change an existing line, but that's a TODO.
104 : ///
105 CBC 9959 : pub fn append(&mut self, option: &str, value: &str) {
106 9959 : self.lines
107 9959 : .push(format!("{}={}\n", option, escape_str(value)));
108 9959 : self.hash.insert(option.to_string(), value.to_string());
109 9959 : }
110 :
111 : /// Append an arbitrary non-setting line to the config file
112 568 : pub fn append_line(&mut self, line: &str) {
113 568 : self.lines.push(line.to_string());
114 568 : }
115 : }
116 :
117 : impl fmt::Display for PostgresConf {
118 : /// Return the whole configuration file as a string
119 568 : fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
120 10527 : for line in self.lines.iter() {
121 10527 : f.write_str(line)?;
122 : }
123 568 : Ok(())
124 568 : }
125 : }
126 :
127 : /// Escape a value for putting in postgresql.conf.
128 9975 : fn escape_str(s: &str) -> String {
129 9975 : // If the string doesn't contain anything that needs quoting or escaping, return it
130 9975 : // as it is.
131 9975 : //
132 9975 : // The first part of the regex, before the '|', matches the INTEGER rule in the
133 9975 : // PostgreSQL flex grammar (guc-file.l). It matches plain integers like "123" and
134 9975 : // "-123", and also accepts units like "10MB". The second part of the regex matches
135 9975 : // the UNQUOTED_STRING rule, and accepts strings that contain a single word, beginning
136 9975 : // with a letter. That covers words like "off" or "posix". Everything else is quoted.
137 9975 : //
138 9975 : // This regex is a bit more conservative than the rules in guc-file.l, so we quote some
139 9975 : // strings that PostgreSQL would accept without quoting, but that's OK.
140 9975 :
141 9975 : static UNQUOTED_RE: Lazy<Regex> =
142 9975 : Lazy::new(|| Regex::new(r"(^[-+]?[0-9]+[a-zA-Z]*$)|(^[a-zA-Z][a-zA-Z0-9]*$)").unwrap());
143 9975 :
144 9975 : if UNQUOTED_RE.is_match(s) {
145 8831 : s.to_string()
146 : } else {
147 : // Otherwise escape and quote it
148 1144 : let s = s
149 1144 : .replace('\\', "\\\\")
150 1144 : .replace('\n', "\\n")
151 1144 : .replace('\'', "''");
152 1144 :
153 1144 : "\'".to_owned() + &s + "\'"
154 : }
155 9975 : }
156 :
157 : /// De-escape a possibly-quoted value.
158 : ///
159 : /// See `DeescapeQuotedString` function in PostgreSQL sources for how PostgreSQL
160 : /// does this.
161 4 : fn deescape_str(s: &str) -> Result<String> {
162 4 : // If the string has a quote at the beginning and end, strip them out.
163 4 : if s.len() >= 2 && s.starts_with('\'') && s.ends_with('\'') {
164 3 : let mut result = String::new();
165 3 :
166 3 : let mut iter = s[1..(s.len() - 1)].chars().peekable();
167 20 : while let Some(c) = iter.next() {
168 18 : let newc = if c == '\\' {
169 8 : match iter.next() {
170 1 : Some('b') => '\x08',
171 1 : Some('f') => '\x0c',
172 2 : Some('n') => '\n',
173 1 : Some('r') => '\r',
174 1 : Some('t') => '\t',
175 2 : Some('0'..='7') => {
176 : // TODO
177 1 : bail!("octal escapes not supported");
178 : }
179 1 : Some(n) => n,
180 UBC 0 : None => break,
181 : }
182 CBC 10 : } else if c == '\'' && iter.peek() == Some(&'\'') {
183 : // doubled quote becomes just one quote
184 1 : iter.next().unwrap()
185 : } else {
186 9 : c
187 : };
188 :
189 17 : result.push(newc);
190 : }
191 2 : Ok(result)
192 : } else {
193 1 : Ok(s.to_string())
194 : }
195 4 : }
196 :
197 1 : #[test]
198 1 : fn test_postgresql_conf_escapes() -> Result<()> {
199 1 : assert_eq!(escape_str("foo bar"), "'foo bar'");
200 : // these don't need to be quoted
201 1 : assert_eq!(escape_str("foo"), "foo");
202 1 : assert_eq!(escape_str("123"), "123");
203 1 : assert_eq!(escape_str("+123"), "+123");
204 1 : assert_eq!(escape_str("-10"), "-10");
205 1 : assert_eq!(escape_str("1foo"), "1foo");
206 1 : assert_eq!(escape_str("foo1"), "foo1");
207 1 : assert_eq!(escape_str("10MB"), "10MB");
208 1 : assert_eq!(escape_str("-10kB"), "-10kB");
209 :
210 : // these need quoting and/or escaping
211 1 : assert_eq!(escape_str("foo bar"), "'foo bar'");
212 1 : assert_eq!(escape_str("fo'o"), "'fo''o'");
213 1 : assert_eq!(escape_str("fo\no"), "'fo\\no'");
214 1 : assert_eq!(escape_str("fo\\o"), "'fo\\\\o'");
215 1 : assert_eq!(escape_str("10 cats"), "'10 cats'");
216 :
217 : // Test de-escaping
218 1 : assert_eq!(deescape_str(&escape_str("foo"))?, "foo");
219 1 : assert_eq!(deescape_str(&escape_str("fo'o\nba\\r"))?, "fo'o\nba\\r");
220 1 : assert_eq!(deescape_str("'\\b\\f\\n\\r\\t'")?, "\x08\x0c\n\r\t");
221 :
222 : // octal-escapes are currently not supported
223 1 : assert!(deescape_str("'foo\\7\\07\\007'").is_err());
224 :
225 1 : Ok(())
226 1 : }
|
