diff options
author | Neil <nyamatongwe@gmail.com> | 2013-07-01 16:37:06 +1000 |
---|---|---|
committer | Neil <nyamatongwe@gmail.com> | 2013-07-01 16:37:06 +1000 |
commit | 92bd3a82d73dd2ea22df882b0b62e7f5878e5420 (patch) | |
tree | af35507a7c121fae3916a4f69c357f163927248c /src/LexGen.py | |
parent | c47b81b98f01cf84f5559adae256bff58b54005f (diff) | |
download | scintilla-mirror-92bd3a82d73dd2ea22df882b0b62e7f5878e5420.tar.gz |
Move non-platform-specific scripts into the scripts directory.
Use FileGenerator module for file generation instead of code in each script.
Diffstat (limited to 'src/LexGen.py')
-rwxr-xr-x | src/LexGen.py | 408 |
1 files changed, 0 insertions, 408 deletions
diff --git a/src/LexGen.py b/src/LexGen.py deleted file mode 100755 index 579468abe..000000000 --- a/src/LexGen.py +++ /dev/null @@ -1,408 +0,0 @@ -#!/usr/bin/env python -# LexGen.py - implemented 2002 by Neil Hodgson neilh@scintilla.org -# Released to the public domain. - -# Regenerate the Scintilla and SciTE source files that list -# all the lexers and all the properties files. -# Should be run whenever a new lexer is added or removed. -# Requires Python 2.4 or later -# Most files are regenerated in place with templates stored in comments. -# The VS .NET project file is generated into a different file as the -# VS .NET environment will not retain comments when modifying the file. -# The files are copied to a string apart from sections between a -# ++Autogenerated comment and a --Autogenerated comment which is -# generated by the CopyWithInsertion function. After the whole -# string is instantiated, it is compared with the target file and -# if different the file is rewritten. -# Does not regenerate the Visual C++ 6 project files but does the VS .NET -# project file. - -import string -import sys -import os -import glob -import codecs -import datetime - -# EOL constants -CR = "\r" -LF = "\n" -CRLF = "\r\n" -if sys.platform == "win32": - NATIVE = CRLF -else: - # Yes, LF is the native EOL even on Mac OS X. CR is just for - # Mac OS <=9 (a.k.a. "Mac Classic") - NATIVE = LF - -# Automatically generated sections contain start and end comments, -# a definition line and the results. -# The results are replaced by regenerating based on the definition line. -# The definition line is a comment prefix followed by "**". -# If there is a digit after the ** then this indicates which list to use -# and the digit and next character are not part of the definition -# Backslash is used as an escape within the definition line. -# The part between \( and \) is repeated for each item in the list. -# \* is replaced by each list item. \t, and \n are tab and newline. -def CopyWithInsertion(input, commentPrefix, retainDefs, eolType, *lists): - copying = 1 - listid = 0 - output = [] - for line in input.splitlines(0): - isStartGenerated = line.startswith(commentPrefix + "++Autogenerated") - if copying and not isStartGenerated: - output.append(line) - if isStartGenerated: - if retainDefs: - output.append(line) - copying = 0 - definition = "" - elif not copying and line.startswith(commentPrefix + "**"): - if retainDefs: - output.append(line) - definition = line[len(commentPrefix + "**"):] - if (commentPrefix == "<!--") and (" -->" in definition): - definition = definition.replace(" -->", "") - listid = 0 - if definition[0] in string.digits: - listid = int(definition[:1]) - definition = definition[2:] - # Hide double slashes as a control character - definition = definition.replace("\\\\", "\001") - # Do some normal C style transforms - definition = definition.replace("\\n", "\n") - definition = definition.replace("\\t", "\t") - # Get the doubled backslashes back as single backslashes - definition = definition.replace("\001", "\\") - startRepeat = definition.find("\\(") - endRepeat = definition.find("\\)") - intro = definition[:startRepeat] - out = "" - if intro.endswith("\n"): - pos = 0 - else: - pos = len(intro) - out += intro - middle = definition[startRepeat+2:endRepeat] - for i in lists[listid]: - item = middle.replace("\\*", i) - if pos and (pos + len(item) >= 80): - out += "\\\n" - pos = 0 - out += item - pos += len(item) - if item.endswith("\n"): - pos = 0 - outro = definition[endRepeat+2:] - out += outro - out = out.replace("\n", eolType) # correct EOLs in generated content - output.append(out) - elif line.startswith(commentPrefix + "--Autogenerated"): - copying = 1 - if retainDefs: - output.append(line) - output = [line.rstrip(" \t") for line in output] # trim trailing whitespace - return eolType.join(output) + eolType - -def UpdateFile(filename, updated): - """ If the file is different to updated then copy updated - into the file else leave alone so CVS and make don't treat - it as modified. """ - try: - infile = open(filename, "rb") - except IOError: # File is not there yet - out = open(filename, "wb") - out.write(updated.encode('utf-8')) - out.close() - print("New %s" % filename) - return - original = infile.read() - infile.close() - original = original.decode('utf-8') - if updated != original: - os.unlink(filename) - out = open(filename, "wb") - out.write(updated.encode('utf-8')) - out.close() - print("Changed %s " % filename) - #~ else: - #~ print "Unchanged", filename - -def Generate(inpath, outpath, commentPrefix, eolType, *lists): - """Generate 'outpath' from 'inpath'. - - "eolType" indicates the type of EOLs to use in the generated - file. It should be one of following constants: LF, CRLF, - CR, or NATIVE. - """ - #print "generate '%s' -> '%s' (comment prefix: %r, eols: %r)"\ - # % (inpath, outpath, commentPrefix, eolType) - try: - infile = open(inpath, "rb") - except IOError: - print("Can not open %s" % inpath) - return - original = infile.read() - infile.close() - original = original.decode('utf-8') - updated = CopyWithInsertion(original, commentPrefix, - inpath == outpath, eolType, *lists) - UpdateFile(outpath, updated) - -def Regenerate(filename, commentPrefix, eolType, *lists): - """Regenerate the given file. - - "eolType" indicates the type of EOLs to use in the generated - file. It should be one of following constants: LF, CRLF, - CR, or NATIVE. - """ - Generate(filename, filename, commentPrefix, eolType, *lists) - -def FindModules(lexFile): - modules = [] - f = open(lexFile) - for l in f.readlines(): - if l.startswith("LexerModule"): - l = l.replace("(", " ") - modules.append(l.split()[1]) - return modules - -# Properties that start with lexer. or fold. are automatically found but there are some -# older properties that don't follow this pattern so must be explicitly listed. -knownIrregularProperties = [ - "fold", - "styling.within.preprocessor", - "tab.timmy.whinge.level", - "asp.default.language", - "html.tags.case.sensitive", - "ps.level", - "ps.tokenize", - "sql.backslash.escapes", - "nsis.uservars", - "nsis.ignorecase" -] - -def FindProperties(lexFile): - properties = {} - f = open(lexFile) - for l in f.readlines(): - if ("GetProperty" in l or "DefineProperty" in l) and "\"" in l: - l = l.strip() - if not l.startswith("//"): # Drop comments - propertyName = l.split("\"")[1] - if propertyName.lower() == propertyName: - # Only allow lower case property names - if propertyName in knownIrregularProperties or \ - propertyName.startswith("fold.") or \ - propertyName.startswith("lexer."): - properties[propertyName] = 1 - return properties - -def FindPropertyDocumentation(lexFile): - documents = {} - f = open(lexFile) - name = "" - for l in f.readlines(): - l = l.strip() - if "// property " in l: - propertyName = l.split()[2] - if propertyName.lower() == propertyName: - # Only allow lower case property names - name = propertyName - documents[name] = "" - elif "DefineProperty" in l and "\"" in l: - propertyName = l.split("\"")[1] - if propertyName.lower() == propertyName: - # Only allow lower case property names - name = propertyName - documents[name] = "" - elif name: - if l.startswith("//"): - if documents[name]: - documents[name] += " " - documents[name] += l[2:].strip() - elif l.startswith("\""): - l = l[1:].strip() - if l.endswith(";"): - l = l[:-1].strip() - if l.endswith(")"): - l = l[:-1].strip() - if l.endswith("\""): - l = l[:-1] - # Fix escaped double quotes - l = l.replace("\\\"", "\"") - documents[name] += l - else: - name = "" - for name in list(documents.keys()): - if documents[name] == "": - del documents[name] - return documents - -def ciCompare(a,b): - return cmp(a.lower(), b.lower()) - -def ciKey(a): - return a.lower() - -def sortListInsensitive(l): - try: # Try key function - l.sort(key=ciKey) - except TypeError: # Earlier version of Python, so use comparison function - l.sort(ciCompare) - -def UpdateLineInFile(path, linePrefix, lineReplace): - lines = [] - updated = False - with codecs.open(path, "r", "utf-8") as f: - for l in f.readlines(): - l = l.rstrip() - if not updated and l.startswith(linePrefix): - lines.append(lineReplace) - updated = True - else: - lines.append(l) - contents = NATIVE.join(lines) + NATIVE - UpdateFile(path, contents) - -host = "prdownloads.sourceforge.net/" -def UpdateDownloadLinks(path, version): - lines = [] - with open(path, "r") as f: - for l in f.readlines(): - l = l.rstrip() - if host in l: - start, prd, rest = l.partition(host) - pth, dot, ending = rest.partition(".") - pthNew = pth[:-3] + version.rstrip() - lineWithNewVersion = start + prd +pthNew + dot + ending - lines.append(lineWithNewVersion) - else: - lines.append(l) - contents = NATIVE.join(lines) + NATIVE - UpdateFile(path, contents) - -def UpdateVersionNumbers(root): - with open(root + "scintilla/version.txt") as f: - version = f.read() - versionDotted = version[0] + '.' + version[1] + '.' + version[2] - versionCommad = version[0] + ', ' + version[1] + ', ' + version[2] + ', 0' - with open(root + "scintilla/doc/index.html") as f: - dateModified = [l for l in f.readlines() if "Date.Modified" in l][0].split('\"')[3] - # 20130602 - # index.html, SciTE.html - dtModified = datetime.datetime.strptime(dateModified, "%Y%m%d") - yearModified = dateModified[0:4] - monthModified = dtModified.strftime("%B") - dayModified = "%d" % dtModified.day - mdyModified = monthModified + " " + dayModified + " " + yearModified - # May 22 2013 - # index.html, SciTE.html - dmyModified = dayModified + " " + monthModified + " " + yearModified - # 22 May 2013 - # ScintillaHistory.html -- only first should change - myModified = monthModified + " " + yearModified - # scite/src/SciTE.h - #define COPYRIGHT_DATES "December 1998-May 2013" - #define COPYRIGHT_YEARS "1998-2013" - dateLine = f.readlines() - - UpdateLineInFile(root + "scintilla/win32/ScintRes.rc", "#define VERSION_SCINTILLA", - "#define VERSION_SCINTILLA \"" + versionDotted + "\"") - UpdateLineInFile(root + "scintilla/win32/ScintRes.rc", "#define VERSION_WORDS", - "#define VERSION_WORDS " + versionCommad) - UpdateLineInFile(root + "scintilla/qt/ScintillaEditBase/ScintillaEditBase.pro", - "VERSION =", - "VERSION = " + versionDotted) - UpdateLineInFile(root + "scintilla/qt/ScintillaEdit/ScintillaEdit.pro", - "VERSION =", - "VERSION = " + versionDotted) - UpdateLineInFile(root + "scintilla/doc/ScintillaDownload.html", " Release", - " Release " + versionDotted) - UpdateDownloadLinks(root + "scintilla/doc/ScintillaDownload.html", version) - UpdateLineInFile(root + "scintilla/doc/index.html", - ' <font color="#FFCC99" size="3"> Release version', - ' <font color="#FFCC99" size="3"> Release version ' + versionDotted + '<br />') - UpdateLineInFile(root + "scintilla/doc/index.html", - ' Site last modified', - ' Site last modified ' + mdyModified + '</font>') - UpdateLineInFile(root + "scintilla/doc/ScintillaHistory.html", - ' Released ', - ' Released ' + dmyModified + '.') - - if os.path.exists(root + "scite"): - UpdateLineInFile(root + "scite/src/SciTE.h", "#define VERSION_SCITE", - "#define VERSION_SCITE \"" + versionDotted + "\"") - UpdateLineInFile(root + "scite/src/SciTE.h", "#define VERSION_WORDS", - "#define VERSION_WORDS " + versionCommad) - UpdateLineInFile(root + "scite/src/SciTE.h", "#define COPYRIGHT_DATES", - '#define COPYRIGHT_DATES "December 1998-' + myModified + '"') - UpdateLineInFile(root + "scite/src/SciTE.h", "#define COPYRIGHT_YEARS", - '#define COPYRIGHT_YEARS "1998-' + yearModified + '"') - UpdateLineInFile(root + "scite/doc/SciTEDownload.html", " Release", - " Release " + versionDotted) - UpdateDownloadLinks(root + "scite/doc/SciTEDownload.html", version) - UpdateLineInFile(root + "scite/doc/SciTE.html", - ' <font color="#FFCC99" size="3"> Release version', - ' <font color="#FFCC99" size="3"> Release version ' + versionDotted + '<br />') - UpdateLineInFile(root + "scite/doc/SciTE.html", - ' Site last modified', - ' Site last modified ' + mdyModified + '</font>') - UpdateLineInFile(root + "scite/doc/SciTE.html", - ' <meta name="Date.Modified"', - ' <meta name="Date.Modified" content="' + dateModified + '" />') - -def RegenerateAll(): - root="../../" - - # Find all the lexer source code files - lexFilePaths = glob.glob(root + "scintilla/lexers/Lex*.cxx") - sortListInsensitive(lexFilePaths) - lexFiles = [os.path.basename(f)[:-4] for f in lexFilePaths] - print(lexFiles) - lexerModules = [] - lexerProperties = {} - propertyDocuments = {} - for lexFile in lexFilePaths: - lexerModules.extend(FindModules(lexFile)) - for k in FindProperties(lexFile).keys(): - lexerProperties[k] = 1 - documents = FindPropertyDocumentation(lexFile) - for k in documents.keys(): - if k not in propertyDocuments: - propertyDocuments[k] = documents[k] - sortListInsensitive(lexerModules) - lexerProperties = list(lexerProperties.keys()) - sortListInsensitive(lexerProperties) - - # Generate HTML to document each property - # This is done because tags can not be safely put inside comments in HTML - documentProperties = list(propertyDocuments.keys()) - sortListInsensitive(documentProperties) - propertiesHTML = [] - for k in documentProperties: - propertiesHTML.append("\t<tr id='property-%s'>\n\t<td>%s</td>\n\t<td>%s</td>\n\t</tr>" % - (k, k, propertyDocuments[k])) - - # Find all the SciTE properties files - otherProps = ["abbrev.properties", "Embedded.properties", "SciTEGlobal.properties", "SciTE.properties"] - if os.path.exists(root + "scite"): - propFilePaths = glob.glob(root + "scite/src/*.properties") - sortListInsensitive(propFilePaths) - propFiles = [os.path.basename(f) for f in propFilePaths if os.path.basename(f) not in otherProps] - sortListInsensitive(propFiles) - print(propFiles) - - Regenerate(root + "scintilla/src/Catalogue.cxx", "//", NATIVE, lexerModules) - Regenerate(root + "scintilla/win32/scintilla.mak", "#", NATIVE, lexFiles) - if os.path.exists(root + "scite"): - Regenerate(root + "scite/win32/makefile", "#", NATIVE, propFiles) - Regenerate(root + "scite/win32/scite.mak", "#", NATIVE, propFiles) - Regenerate(root + "scite/src/SciTEProps.cxx", "//", NATIVE, lexerProperties) - Regenerate(root + "scite/doc/SciTEDoc.html", "<!--", NATIVE, propertiesHTML) - Generate(root + "scite/boundscheck/vcproj.gen", - root + "scite/boundscheck/SciTE.vcproj", "#", NATIVE, lexFiles) - - UpdateVersionNumbers(root) - -RegenerateAll() |