| cursor |
|
More...
|
about 17 years ago
|
| counter-reset |
|
More...
|
about 17 years ago
|
| counter-increment |
|
More...
|
about 17 years ago
|
| Factor out common part of content parser. This simplifies and shortens the code. |
|
More...
|
about 17 years ago
|
| Finish content |
|
More...
|
about 17 years ago
|
| Factor out parsing of the value part of list-style-type. Use this when parsing counter functions. |
|
More...
|
about 17 years ago
|
| Changes to the encoding of content in bytecode to better match the spec. Beginnings of a content property parser. |
|
More...
|
about 17 years ago
|
| Something approximating clip |
|
More...
|
about 17 years ago
|
| Rather less memory corruption. |
|
More...
|
about 17 years ago
|
| border-spacing |
|
More...
|
about 17 years ago
|
| Fix up azimuth to cater for !important after property value. Something approximating background-position. |
|
More...
|
about 17 years ago
|
| Finish azimuth |
|
More...
|
about 17 years ago
|
| The advantage of interning strings is that you don't have to store their length everywhere. Purge the length part from the encoding of a string in the bytecode. Fix bytecode dump code to cope with this. |
|
More...
|
about 17 years ago
|
| Given that we're going to be copying selector details, anyway, it seems pointless to create objects on the heap which can equally well be on the stack, given the shortness of their lifetime. |
|
More...
|
about 17 years ago
|
| Retain pointer to dictionary entries so we don't have to rediscover it later. |
|
More...
|
about 17 years ago
|
| Simplify decision as to whether to intern token data by inserting markers into the css_token_type enum. |
|
More...
|
about 17 years ago
|
| Divorce css_string from whatever gets stored in lpu hashes. Use pointers to parserutils_hash_entry directly in stylesheet datastructures. |
|
More...
|
about 17 years ago
|
| More modifications to allow numerous iterations for profiling. Overall time breakdown for allzengarden.css is approximately: |
|
More...
|
about 17 years ago
|
| Modify to allow a configurable number of iterations for profiling. |
|
More...
|
about 17 years ago
|
| Make getToken reduce consecutive whitespace tokens to a single whitespace token. Update eatWS appropriately. This reduces the number of calls to getToken by a million or so. |
|
More...
|
about 17 years ago
|
| Slightly clearer code. Marginally faster, too. |
|
More...
|
about 17 years ago
|
| Use parserutils_hash instead of parserutils_dict. This approximately halves the size of the interned string table. We now have the following for allzengarden.css: |
|
More...
|
about 17 years ago
|
| Commentary on which tokens actually need string data interning. |
|
More...
|
about 17 years ago
|
| Beginnings of azimuth. |
|
More...
|
about 17 years ago
|
| Tidy things up somewhat. css21 is now language, as everything will share the same parsing rules (although there is facility to alter behaviour based upon the language level -- consult language->sheet->level and then decide what to do) |
|
More...
|
about 17 years ago
|
| Optimise css_stylesheet_add_rule by the cunning approach of not iterating through a singly linked list to find the end every time we want to insert a rule. This doubles parsing speed. |
|
More...
|
about 17 years ago
|
| Change the way in which css_rules are defined. This is more compact than the previous approach. Space requirements for rule objects alone is now 447,120 bytes as opposed to 819,270 bytes previously. This reduces the space requirements for allzengarden.css to 2,041,712 bytes, which is 45% the size of what we had originally. |
|
More...
|
about 17 years ago
|
| css_string is now the same as a parserutils_dict_entry. This allows us to use dict entries directly as strings. |
|
More...
|
about 17 years ago
|
| Calculate some kind of ballpark figure for the memory requirements of a stylesheet. Summary: lots. |
|
More...
|
about 17 years ago
|
| Prevent reading from beyond the start of token data. |
|
More...
|
about 17 years ago
|
