Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | automated conversion of svn:ignore properties to .gitignore files on pullscanpullscan | Adrian Thurston | 2012-04-28 | 12 | -0/+36 |
| | |||||
* | Some cleanup. | Adrian Thurston | 2009-03-09 | 2 | -19/+15 |
| | |||||
* | Some cleanup. | Adrian Thurston | 2009-03-08 | 3 | -11/+15 |
| | |||||
* | Don't need to recompute p when there is traling context in all of send_token, | Adrian Thurston | 2009-03-08 | 1 | -13/+8 |
| | | | | | send_ignore and exec_gen. Can do it after FsmRun::exeute is called in FsmRun::scan. | ||||
* | Fix to last commit. | Adrian Thurston | 2009-03-08 | 1 | -1/+1 |
| | |||||
* | Log the token region in FsmRun::scanToken. | Adrian Thurston | 2009-03-08 | 1 | -19/+7 |
| | |||||
* | Since there is no longer a continually run scanner loop that pushes tokens, we | Adrian Thurston | 2009-03-08 | 2 | -45/+4 |
| | | | | | can initialize FsmRun::cs and FsmRun::tokstart at the beginning of teach run of the scanner in which we attempt to fetch a token. | ||||
* | Converted the token generation and sending functions in fsmRun to non-member | Adrian Thurston | 2009-03-08 | 3 | -119/+100 |
| | | | | | functions so we can avoid calls to fsmRun member functions on the stack when we need to switch the scanning source. | ||||
* | Got parse_stop working again. Don't need to set fsmRun->cs to error anymore | Adrian Thurston | 2009-03-08 | 1 | -15/+14 |
| | | | | because we don't need to break out of the fsmRun loop deeper in the stack. | ||||
* | Scanner error handling code moved to separate function. | Adrian Thurston | 2009-03-08 | 1 | -54/+60 |
| | |||||
* | More work on the pull scanner. Seems to work parsing input and constructors. | Adrian Thurston | 2009-03-08 | 3 | -112/+141 |
| | |||||
* | Took the 'parser' variable out of FsmRun. Changed FsmRun::run to FsmRun::scan | Adrian Thurston | 2009-03-07 | 6 | -55/+84 |
| | | | | and started on a separte scan-parse-repeat loop. | ||||
* | Record the matched token in emitToken and the EMIT_TOKEN code generator. | Adrian Thurston | 2009-03-06 | 4 | -16/+24 |
| | |||||
* | Started on the pull scanner. The FsmRun::execute function returns after | Adrian Thurston | 2009-03-06 | 3 | -16/+16 |
| | | | | matching a token. | ||||
* | Redid the branch for the pull scanner, this time at the root of the project and | Adrian Thurston | 2009-03-06 | 193 | -131/+27383 |
|\ | | | | | | | not the colm subdir. | ||||
| * | Fixed broken build from last commit. | Adrian Thurston | 2009-03-05 | 1 | -1/+1 |
| | | |||||
| * | Fixed open_file(). | Adrian Thurston | 2009-03-05 | 4 | -25/+16 |
| | | |||||
| * | Eliminated PdaRun::run(). We should be pulling tokens from scanner and pushing | Adrian Thurston | 2009-03-04 | 4 | -11/+5 |
| | | | | | | | | | | them to the parser, rather than calling the scanner and pushing to the parser from there. This is a small start on that. | ||||
| * | Bug fix: match_text was getting into the list of locals. It loaded with a | Adrian Thurston | 2009-03-04 | 1 | -1/+3 |
| | | | | | | | | special bytecode and is not on the stack. | ||||
| * | Regex subexpression captures are now bound as local variables in token ↵ | Adrian Thurston | 2009-03-02 | 6 | -13/+64 |
| | | | | | | | | generation actions. | ||||
| * | Added objFieldList so that object attributes have their offets ordered as they | Adrian Thurston | 2009-03-01 | 4 | -15/+48 |
| | | | | | | | | | | | | appear. Otherwise the args to make_token need to be sorted according to the alphabetical sort of the attribute names. Also, debugging gets harder with an alphabetical sort and this has caused me problems on more than one occation. | ||||
| * | Include captured attributes in constructor patterns and add them to the | Adrian Thurston | 2009-03-01 | 2 | -0/+32 |
| | | | | | | | | construced tree at runtime. | ||||
| * | Capture info now propagated to the runtime data. In make token, the LEL's list | Adrian Thurston | 2009-02-28 | 6 | -17/+60 |
| | | | | | | | | | | | | | | of captures is used to create the attributes. No longer relying on the end markers. That scheme was wrong since the index of the end marker can't correctly correspond to the attribute offset, otherwise there could be interference among tokens. | ||||
| * | Some work on unifying the marking facilities for trailing context and sub-regex | Adrian Thurston | 2009-02-27 | 14 | -67/+87 |
| | | | | | | | | | | | | | | capture. Now using the same marking array for the trailing context and the capture. The set markers are now shifted long with partial token matches. But not yet propigating the list of captures associated with a token to the runtime data. | ||||
| * | Got trailing context working. The marks are not yet shifted when the with the | Adrian Thurston | 2009-02-27 | 15 | -116/+94 |
| | | | | | | | | buffer contents. | ||||
| * | Added a sync pattern. | Adrian Thurston | 2009-02-27 | 1 | -0/+2 |
| | | |||||
| * | removed some duplicate position updates | Adrian Thurston | 2009-02-26 | 1 | -2/+3 |
| | | |||||
| * | When the scanner is generated we can decide then if we need to call a token | Adrian Thurston | 2009-02-26 | 3 | -67/+45 |
| | | | | | | | | generation action. Call execGen right from fsmExec. Allows some cleanup. | ||||
| * | Can now have generation code blocks on ignore tokens. The send function now | Adrian Thurston | 2009-02-26 | 2 | -53/+94 |
| | | | | | | | | | | | | checks the ignore property of a token to determine if it should be ignored. This will require making an ignore_def production type to have context-free comments. This will be a normal production but with the ignore bit set. | ||||
| * | The print_xml function now omits attributes and comments. Now use print_xml_ac | Adrian Thurston | 2009-02-26 | 8 | -16/+44 |
| | | | | | | | | to get those in the printout. | ||||
| * | Need to escape '<', '>', and '&'. | Adrian Thurston | 2009-02-25 | 1 | -1/+7 |
| | | |||||
| * | cleaned up a large function | Adrian Thurston | 2009-02-25 | 2 | -113/+125 |
| | | |||||
| * | new precedence keywords | Adrian Thurston | 2009-02-25 | 1 | -1/+1 |
| | | |||||
| * | Added the plus operator. | Adrian Thurston | 2009-02-25 | 9 | -35/+101 |
| | | |||||
| * | Definition names must be unique. | Adrian Thurston | 2009-02-23 | 1 | -5/+6 |
| | | |||||
| * | Grammar definition names must be unique. | Adrian Thurston | 2009-02-23 | 1 | -10/+4 |
| | | |||||
| * | Flattened arg and specified precedence. | Adrian Thurston | 2009-02-22 | 1 | -70/+49 |
| | | |||||
| * | Can now set the precedence of a production as you can in bison. | Adrian Thurston | 2009-02-22 | 4 | -25/+58 |
| | | |||||
| * | implemented nonassoc | Adrian Thurston | 2009-02-22 | 2 | -3/+24 |
| | | |||||
| * | Fix to last commit. | Adrian Thurston | 2009-02-22 | 2 | -5/+7 |
| | | |||||
| * | Started precedence implementation. | Adrian Thurston | 2009-02-22 | 6 | -4/+130 |
| | | |||||
| * | The getKlangEl now always creates tokens with the Unknown tpe. If it comes back | Adrian Thurston | 2009-02-16 | 4 | -43/+36 |
| | | | | | | | | | | | | unknown then it hasn't been previously defined as anything. Now reporting an error if tokens, maps and lists are defined more than once. | ||||
| * | Don't allow KW_Def statements to add to a the production list. This is left ↵ | Adrian Thurston | 2009-02-16 | 1 | -34/+32 |
| | | | | | | | | over from Kelbt. | ||||
| * | Unset action orderings are a reality due to the fact that we are using LALR(1). | Adrian Thurston | 2009-02-15 | 1 | -11/+5 |
| | | | | | | | | | | They are not a user-level warning. Just ordering last since they result in parse errors anyways. | ||||
| * | label the states | Adrian Thurston | 2009-02-15 | 1 | -2/+2 |
| | | |||||
| * | added a couple missing includes | Adrian Thurston | 2009-02-15 | 1 | -0/+2 |
| | | |||||
| * | Parse errors in patterns and constructors now report the line number in the | Adrian Thurston | 2009-02-15 | 8 | -41/+69 |
| | | | | | | | | space of the colm source file. | ||||
| * | A couple fixes to the last patch. | Adrian Thurston | 2009-02-13 | 2 | -3/+6 |
| | | |||||
| * | Now using a single parse table with multiple entry points, instead of multiple | Adrian Thurston | 2009-02-13 | 14 | -154/+220 |
| | | | | | | | | | | | | | | disjoint tables, one for each nonterminal that must be parsed. I had pursued this before but gave up because the EOF token of one parser could interfere with another parser. The observation I was missing was that we can use unique EOF tokens for each nonterminal we parse, thus avoiding interference. | ||||
| * | got graphviz output working | Adrian Thurston | 2009-02-08 | 6 | -304/+62 |
| | |