| 1 |
# 0.3.5 (31/07/2024) |
| 2 |
|
| 3 |
The version 0.3.5 is a bugfix and enhancement release for the 0.3 branch. |
| 4 |
|
| 5 |
## Changes summary |
| 6 |
|
| 7 |
* Added ctpl_value_take_string() to reduce memory copies, and used it |
| 8 |
internally where appropriate; |
| 9 |
* Fixed build with GCC 14; |
| 10 |
|
| 11 |
|
| 12 |
# 0.3.4 (03/02/2015) |
| 13 |
|
| 14 |
The version 0.3.4 is a bugfix and enhancement release for the 0.3 branch. |
| 15 |
|
| 16 |
## Changes summary |
| 17 |
|
| 18 |
* Added Windows support to the command-line utility (Andrea Zagli); |
| 19 |
* Added French and Italian translations; |
| 20 |
* Fixed various build system issues; |
| 21 |
* Added support for building for Windows with Automake; |
| 22 |
|
| 23 |
|
| 24 |
# 0.3.3 (11/08/2011) |
| 25 |
|
| 26 |
The version 0.3.3 is a bugfix release for the 0.3 branch. |
| 27 |
|
| 28 |
## Changes summary |
| 29 |
|
| 30 |
* Fixed crash when indexing expressions; |
| 31 |
* Fixed two possible reads of uninitialized memory; |
| 32 |
* Plugged a few memory leaks; |
| 33 |
|
| 34 |
|
| 35 |
# 0.3.2 (07/01/2011) |
| 36 |
|
| 37 |
The version 0.3.2 is a bugfix and enhancement release for the 0.3 branch. |
| 38 |
|
| 39 |
## Changes summary |
| 40 |
|
| 41 |
* Fixed crash comparing arrays with other types; |
| 42 |
* Actually implemented array comparison; |
| 43 |
* Added support for string literals in expressions; |
| 44 |
* Fixed wrong reading of some numbers that led to parsing failure when the |
| 45 |
addition or subtraction operators was not surrounded by blank characters. |
| 46 |
This bug was introduced in version 0.3.1. |
| 47 |
|
| 48 |
|
| 49 |
# 0.3.1 (13/11/2010) |
| 50 |
|
| 51 |
The version 0.3.1 is a bugfix and enhancement release for the 0.3 branch. |
| 52 |
|
| 53 |
## Changes summary |
| 54 |
|
| 55 |
* Fixed crash in addition operator when handling arrays; |
| 56 |
* Improved floating-point numbers output; |
| 57 |
* Added support for blanks between a symbol and its index; |
| 58 |
* Added support for empty arrays in environment descriptions; |
| 59 |
|
| 60 |
The most interesting change of this release is probably the improved output of |
| 61 |
floating-points numbers. Floating point numbers always suffered from false |
| 62 |
precision: for example "3.14" was wrongly rendered as "3.1400000000000001". |
| 63 |
This problem should be fixed now. |
| 64 |
|
| 65 |
### Internal changes that you might notice |
| 66 |
|
| 67 |
The way numbers are read has changed a little bit, now only the characters |
| 68 |
that forms a fully valid number are read. This may lead to a different error |
| 69 |
with some malformed templates -- even though there will still be an error, but |
| 70 |
maybe at another step of the lexing process. |
| 71 |
|
| 72 |
For example, template "{0x}" raised an error about the number lacking a |
| 73 |
mantissa (0x is a valid prefix for an hexadecimal value, but lacks the actual |
| 74 |
hexadecimal number), and now correctly read the number but raises an error |
| 75 |
about character "x" being unexpected. |
| 76 |
|
| 77 |
|
| 78 |
# 0.3 (09/10/2010) |
| 79 |
|
| 80 |
The version 0.3 is an API and ABI break from the previous 0.2 branch. |
| 81 |
|
| 82 |
## Changes summary |
| 83 |
|
| 84 |
* Heavy API cleanup; |
| 85 |
* Array expressions can now be indexed; |
| 86 |
* The `for` loop now iterates over an expression rather than a symbol name; |
| 87 |
* CtplEnviron now uses reference counting; |
| 88 |
* Improvements of CtplEnviron's memory usage; |
| 89 |
* Bug fixes: |
| 90 |
* ctpl_value_copy(), ctpl_value_set_string() and ctpl_value_set_array*() can |
| 91 |
now take a part of itself as the value to set/copy; |
| 92 |
* Fix internal handling of an error which may lead to wrong error detection; |
| 93 |
* Fix wrong name of streams built from GFiles, paths and URIs; |
| 94 |
* Fix dramatic slowness when a symbol gets pushed many times; |
| 95 |
* Some internal refactoring. |
| 96 |
|
| 97 |
Yes, it is the second API break since the library's birth, less than one year |
| 98 |
ago. I know it and I apologize for the inconvenience. |
| 99 |
|
| 100 |
But even if it's an API break, if you make normal usage of the library, it |
| 101 |
shouldn't be too painful for you to upgrade to the new version (see below). |
| 102 |
Many functions and data structures have been removed or made private, but the |
| 103 |
API that was meant to be used by final users had only small changes. |
| 104 |
|
| 105 |
I will also justify this new break, at least for you to understand the |
| 106 |
original motivations for it: |
| 107 |
* There was way too many symbols exposed, and a lot of them wasn't supposed to |
| 108 |
be used by library's users; |
| 109 |
* Exposing data structures (as well as some other internals) makes working on |
| 110 |
the core harder, because any change to one of this data structure means an |
| 111 |
API break. |
| 112 |
Then this new break is here to -- hopefully -- ease future improvements of |
| 113 |
CTPL, as well as preventing more API breaks. |
| 114 |
|
| 115 |
## What to change in your code |
| 116 |
|
| 117 |
### Token-related functions |
| 118 |
|
| 119 |
The functions ctpl_token_free() and ctpl_token_expr_free() lost their second |
| 120 |
(boolean) argument. If you passed TRUE here, simply remove that second |
| 121 |
argument. Otherwise, you did strange things and you're screwed (but see the |
| 122 |
next section). |
| 123 |
|
| 124 |
ctpl_lexer_free_tree() has also been removed, use ctpl_token_free() instead |
| 125 |
(the semantic is exactly the same). |
| 126 |
|
| 127 |
### Environments |
| 128 |
|
| 129 |
CtplEnviron now uses reference counting. It means that you get all the good |
| 130 |
of reference counting, but you also have to replace calls to |
| 131 |
ctpl_environ_free() with ctpl_environ_unref(). Note that ctpl_environ_unref() |
| 132 |
does not support NULL argument. |
| 133 |
|
| 134 |
ctpl_environ_pop() has changed. It now takes a third argument to get the |
| 135 |
poped value and returns a boolean (see the documentation for more details). |
| 136 |
Anyway you should probably not need ctpl_environ_pop() at all. |
| 137 |
|
| 138 |
### Input and output streams |
| 139 |
|
| 140 |
ctpl_input_stream_read_double() and ctpl_input_stream_read_long() are renamed |
| 141 |
to ctpl_input_stream_read_float() and ctpl_input_stream_read_int() for more |
| 142 |
consistency with the rest of the API. |
| 143 |
|
| 144 |
### Others |
| 145 |
|
| 146 |
The environment to pass to ctpl_eval_*() is not constant anymore. This will |
| 147 |
probably not change anything for you, but it's a subtle change. This is to |
| 148 |
allow future operators to affect the environment if needed (which isn't the |
| 149 |
case yet). |
| 150 |
|
| 151 |
## The dark side of the cleanup |
| 152 |
|
| 153 |
Apart the inconvenience of an API break, some "features" of the library may be |
| 154 |
gone -- yes, having many internal exposed may allow unwanted "features". |
| 155 |
First, know that these features weren't planned. Then, feel free to contact |
| 156 |
me, explaining what functionality you miss and why, and I'll probably be happy |
| 157 |
to re-introduce it cleanly for you. |
| 158 |
|
| 159 |
## CLI tool changes |
| 160 |
|
| 161 |
The CLI tool, apart then new features of the template engine, gained support |
| 162 |
for encodings. This brings a new `--encoding` option allowing to specify the |
| 163 |
encoding of the input and output files. The default encoding is the system's |
| 164 |
one. |
| 165 |
|
| 166 |
|
| 167 |
# 0.2.2 (25/05/2010) |
| 168 |
|
| 169 |
The version 0.2.2 is a bugfix release for the 0.2 branch. |
| 170 |
|
| 171 |
## Changes summary |
| 172 |
|
| 173 |
* Fix lexing of a whole expression input (using ctpl_lexer_expr_lex() or |
| 174 |
one of its variants); |
| 175 |
* Fix priority of boolean operators AND and OR; |
| 176 |
* Fix support of operators <= and >=; |
| 177 |
* Better error reporting in the expression lexer; |
| 178 |
* New alternative build system: Waf (from a work by Enrico Tröger, thanks). |
| 179 |
|
| 180 |
|
| 181 |
# 0.2.1 (27/04/2010) |
| 182 |
|
| 183 |
The version 0.2.1 is a bugfix release for the 0.2 branch. |
| 184 |
|
| 185 |
## Changes summary |
| 186 |
|
| 187 |
* Do not export the private API; |
| 188 |
* Little cleaning of the build system. |
| 189 |
|
| 190 |
|
| 191 |
# 0.2 (15/04/2010) |
| 192 |
|
| 193 |
The version 0.2 is an API and ABI break from the previous 0.1 branch. |
| 194 |
|
| 195 |
## Changes summary (in-depth): |
| 196 |
|
| 197 |
### New features and important changes |
| 198 |
|
| 199 |
* GIO replaces LibMB. Impact is: |
| 200 |
* Less memory usage (input template may or may not be loaded into memory); |
| 201 |
* Small performance regression, between 8% to 19% [1] (on extreme cases), |
| 202 |
measured on heavy templates; |
| 203 |
* Any GIO stream may be live-lexed, from local files to remote FTP, Samba, |
| 204 |
DAV, etc. |
| 205 |
* Better lexer error reporting, with line and position marks; |
| 206 |
* Improved numeric constants loading, including: |
| 207 |
* Integers are now loaded as integers, impacting: |
| 208 |
* No more precision loss; |
| 209 |
* Shorter maximum value, but still more than reasonable (C's long: in 64 |
| 210 |
bits machines should be something like from -9223372036854775807 to |
| 211 |
+9223372036854775807, and -2147483647 to +2147483647 on 32 bits |
| 212 |
machines). |
| 213 |
* Support for octal and binary constants via 0o and 0b prefix respectively; |
| 214 |
* Improved token representation, saving about from 19% to 28% of memory use |
| 215 |
for the token tree (no measurable performance impact) (already in 0.1.90); |
| 216 |
* Add support for single-line comments in environ descriptions; |
| 217 |
* Fix support of array values that starts with blanks in environ descriptions; |
| 218 |
* `make distcheck` and out-of-tree build now work; |
| 219 |
* Add man page for the ctpl command-line utility. |
| 220 |
|
| 221 |
As you can see, the main improvements of this release are less memory |
| 222 |
consumption (about 30/40% better!) and GIO usage bringing almost universal |
| 223 |
input. |
| 224 |
|
| 225 |
## Quid of the performances? |
| 226 |
|
| 227 |
Yes, we lost about 8% to 19% [1] of speed. But I decided that the gain worth |
| 228 |
it: GIO allows live streams to be parsed without loading them previously into |
| 229 |
memory, which means that the memory usage is about the same minus about the |
| 230 |
size of the input template. |
| 231 |
Since the parsing of a 140M (28366227 lines) example template took 15.9612s |
| 232 |
before and 17.4756s now (8.67% slower), I thought it was reasonable when we |
| 233 |
gained about 140M of memory. Another test (135M input, 10944000 lines) took |
| 234 |
11.4204s before and 13.3378s now (14.38% slower), but needed about 130M less |
| 235 |
memory. |
| 236 |
|
| 237 |
And still note that these are very heavy templates (yes, 11-28.4M lines!), and |
| 238 |
real world templates are really really smaller. And this is the lexing, which |
| 239 |
is done only once per template, the parsing is not impacted. |
| 240 |
|
| 241 |
Finally, note that invalid templates may be found invalid without reading the |
| 242 |
whole input, especially if the invalid data is at the start. This makes for |
| 243 |
example a 350M invalid template to be immediately (huh, about 0.009s, and this |
| 244 |
includes program launching) reported as invalid, and this about 99% faster |
| 245 |
than before (without any I/O latency). |
| 246 |
|
| 247 |
[1] note that the measurement was done with all the changes (except token tree |
| 248 |
optimization), which can mean that the overhead is not only due to GIO but |
| 249 |
to the new features too. |
| 250 |
Another thing to note is that this measurement always used a cached file |
| 251 |
input, meaning that any I/O gain or loss due not to load the whole |
| 252 |
template at once was not took into account; but I/O doesn't make so much |
| 253 |
sense since it depends heavily on the input location (even two hard drives |
| 254 |
might behave quite differently, depending on their speed, cache size and |
| 255 |
so on). But with I/O, results moves from +18.66% to -3.59% slower. |
| 256 |
|
| 257 |
|
| 258 |
# 0.1.90 (06/04/2010) |
| 259 |
|
| 260 |
The version 0.1.90 is the last of the 0.1.x branch. It is the 0.2-minus-one |
| 261 |
release, with a part of the 0.1-compatible improvements that will come in the |
| 262 |
upcoming 0.2 branch. |
| 263 |
|
| 264 |
## Changes summary |
| 265 |
|
| 266 |
* Improved token representation, saving about from 20% to 30% of memory use |
| 267 |
for the token tree (no measurable performance impact); |
| 268 |
* New ctpl_environ_foreach() and ctpl_environ_merge() functions; |
| 269 |
* Fix missing inclusion of readutils.h in the global header. |
| 270 |
|
| 271 |
|
| 272 |
# 0.1.1 (15/03/2010) |
| 273 |
|
| 274 |
The version 0.1.1 is a bugfix release for the 0.1 branch. |
| 275 |
|
| 276 |
## Changes summary |
| 277 |
|
| 278 |
* Fixed stack overflow with very large token trees; |
| 279 |
* Great speed improvement with large token trees. |
| 280 |
|
| 281 |
|
| 282 |
# 0.1 (30/11/2009) |
| 283 |
|
| 284 |
The version 0.1 is the first release of CTPL. CTPL is a template engine |
| 285 |
library aiming to easily allow template parsing from C programs. |