compilers-aho/03-lexical-analysis.md
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
--- geometry: - lmargin=0.9in - rmargin=0.3in - tmargin=0.3in - bmargin=0.5in - twoside papersize: A4 ... \begin{huge} \textbf{Chapter 3 - Lexical Analysis} \end{huge} We usually use *lexical-analyser generators* to which we feed the patterns of the lexemes and the generator then produces code that works as a lexical analyser. These patterns are specified using regular expressions. These expressions are converted into NDFSM and then to DFSM. These two models are then fed to a "driver" that simulates these automaton and decide the next token. # The Role of the Lexical Analyzer - Main task - read the input characters of the source program, group them into lexemes and produces as output a sequence of tokens for each lexeme in the source program. The stream of tokens is then sent to the parser for syntax analysis - It also interacts with the symbol table. - The interactions between the parser and the lexical analyser are depicted in the Fig 1 and it's implemented in a way where the parser calls the lexical analyser by the *getNextToken* command. The calls causes the lexical analyser to read characters from the input and determine the next lexeme and produce the next token for the parser. ![Interaction between Lexical Analyser and parser](img/interaction-parser-lex.png){width=70%} - Lexical analyser may also perform some other tasks like stripping out whitespaces and comments. - It may also indicate errors by inserting error message in the appropriate lines by keeping track of line numbers and may also perform macro expansion. - Sometimes they are divided into **two processes**: 1. **Scanning** consists of simple process that do not require tokenization like deleting comments and whitespaces. 2. **Lexical Analysis** is the main part where the scanner produces output as a sequence of tokens. ## Lexical Analysis Versus Parsing. Reasons why the analysis process of compiler is split into lexical analyser and parser (syntax analyser): 1. **Simplicity of design**. The separation of tasks helps us simplify at leasts one of those tasks. Like the lexical analyser once done with dealing with whitespaces and comments, it's easier and simpler for the syntax analyser to parse it with the assumption that the comments and the whitespaces have been removed rather than having to process them as well. 2. **Compiler efficient is improved.** A separate lexical analyzer helps us to apply specialized techniques to improve the efficiency only of the lexical task. Like one example is using specialized buffering techniques during reading the input to speed up the compiler. 3. **Portability is enhanced**. Input-device-specific peculiarities can be restricted to lexical analyzer. ## Tokens, Patterns, and Lexemes - A **token** is a pair consisting of a token name and a token attribute value. The token name is an abstract symbol representing the kind of lexical unit, e.g., a keyword, or an identifier. - A **pattern** is a description of the form the lexemes of a token may take. In the case of a keyword as a token, the pattern is just a sequence of characters that form that keyword. - A **lexeme** is a sequence of characters that matches the pattern for a token and is identified by the lexical analyzer as an *instance* of that token. ## Attributes for tokens - Attribute values for tokens are used to provide more information about the token. For example a token of number matches both the lexemes `0` and `1`, thus to provide more information to the other phases of the compiler, attribute values are used. - The token name influences how the parsing is done, while the attribute value influences the translation of the tokens after the parse. - The attribute value of ta token may consist a great deal of information about that token. An attribute value can be a structure of information about that token. For example for an **id** token, there can be information about it's type, it's lexeme, or it's location in the source code. All this information is kept in the symbol table. Hence the attribute value is a pointer to an entry in the symbol table. - **Example**: Token names and associated attribute values: E = M * C ** 2 $\langle\textbf{id}, \text{pointer to symbol-table entry for \texttt{E}}\rangle$\ $\langle\textbf{assign\_op}\rangle$\ $\langle\textbf{id}, \text{pointer to symbol-table entry for \texttt{M}}\rangle$\ $\langle\textbf{mult\_op}\rangle$\ $\langle\textbf{id}, \text{pointer to symbol-table entry for \texttt{C}}\rangle$\ $\langle\textbf{exp\_op}\rangle$\ $\langle\textbf{number}, \text{integer value 2}\rangle$\ For the last the token (**number**) a typical compiler would store a character string in the symbol table representing the number. In this case "2" and the pointer to it would be the attribute value of the token. ## Lexical Errors - The lexical analyzer is not always able to identify an error without the help of another component. For example `fi ( a == f(x))...`. In this case `fi` could be a **id** of a function that is not yet declared or it could be a misspelled form of "if". The lexical analyzer pases it as an id and lets the parser decide whether it's valid or not - Lexical errors can be corrected/recovered by the lexical analyzer itself. If no patterns match the prefix of the remaining input, then the lexical analyzer can delete the successive characters from the input until a token is found. - There are a few more such **transformation** that the lexical analyzer can perform to "repair" the input: 1. Delete one character from remaining input. 2. Insert a missing character from remaining input. 3. Replace a character with another character. 4. Transpose two adjacent characters. - A simple strategy may be to see if the remaining input can be transformed into a valid lexeme with a single transformation. # Input Buffering Sometimes it's necessary to lookahead one or more character beyond the next lexeme before we can be sure we have the right lexeme. For example we cannot be sure we've seen the end of an identifier until we see a character that is not a digit or a letter, and therefore is not a part of the identifier. A single character operator like `<`, `=` or `-` could also be the start of a two-character operator like `<=`, `==` or `->`. ## Buffer Pairs - Buffering is used to reduce the amount of overhead required in reading a single character. An important scheme involves two buffers that are alternatively reloaded. ![Pair of input bufferes](img/buffer.png){width=75%} - Each buffer is of the size *N* where *N* is usually the size of a disk block. - We can use one system read command to read *N* characters into the buffer rather than using one system call for each character. - **eof** marks the end of the source program. - Two pointers are maintained: 1. `lexemeBegin`, marks the beginning of the current lexeme whose extent is being determined. 2. `forward` scans ahead until a pattern is found. - When `forward` reaches the end of the buffer, we reload the second buffer and then point forward to the start of the second buffer. - We never overwrite a lexeme in the buffer before determining it. ## Sentinels - If we use the previous scheme we have to make two tests every time we read a character, one for determining the end of buffer and the other to determine what character was read. - We can combine both of these tests using a sentinel character that marks the end of the buffer. The sentinel character is a special character that cannot be a part of the source program. - **eof** is the choice here. It still retains it's use as the end of file marker, as if it's encountered anywhere else other than the end of the buffer it means that we have reached the end of the source program. - Algorithm: ``` switch (*forward++) { case eof: if (forward is at the end of first buffer) { reload second buffer; forward = beginning of second buffer; } else if (forward is at the end of second buffer) { reload first buffer; forward = beginning of first buffer; } else /* eof within a buffer marks the end of input */ terminate lexical analysis; break; /*Cases for other character*/ /* .. */ /* .. */ } ``` # Specification of Tokens ## Strings and Languages ## Operations on Languages ## Regular Expressions - Notation used to describe a language. - Some conventions: 1. The unary operator `*` has highest precedence and is left associative. 2. Concatenations has second highest precedence and is left associative. 3. `|` has lowest precedence and is left associative. - Language that can be defined by a regular expression is called a regular set. if two regular expression $r$ and $s$ denote the same regular set, we say they are equivalent write r=s ![Example of regular expression](img/eg3.4.png){width=70%}# ## Regular Definition - For convenience, we may give names to the regular expressions and use those names in subsequent expressions, as if the names were themselves symbols. - If $\sum$ is the alphabet of basic symbols, then *regular definition* is a sequence of definitions of the form: $$d_1 \to r_1$$ $$d_2 \to r_2$$ $$d_3 \to r_3$$ where: 1. Each $d_i$ is a new symbol, not in $\sum$ and not same as any other *d's* 2. Each $r_i$ is a regular expression over the alphabet $\sum \cup \{d_1, d_2,...,d_{i-1}\}$ - _We replace the uses of $d_1$ in $r_2$ by $r_1$, then replacing the use of $d_1$ and $d_2$ in $r_3$ by $r_1$ and (the substituted) $r_2$ and so on. ![C identifiers regular definition exampele](img/cid-reg-definitions.png){width=40%}