Important changes to repositories hosted on mbed.com
Mbed hosted mercurial repositories are deprecated and are due to be permanently deleted in July 2026.
To keep a copy of this software download the repository Zip archive or clone locally using Mercurial.
It is also possible to export all your personal repositories from the account settings page.
Fork of mbed-sdk-tools by
Diff: arm_pack_manager/__init__.py
- Revision:
- 32:8ea194f6145b
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/arm_pack_manager/__init__.py Wed Jan 04 11:58:24 2017 -0600 @@ -0,0 +1,436 @@ +from urllib2 import urlopen, URLError +from bs4 import BeautifulSoup +from os.path import join, dirname, basename +from os import makedirs +from errno import EEXIST +from threading import Thread +from Queue import Queue +from re import compile, sub +from sys import stderr, stdout +from fuzzywuzzy import process +from itertools import takewhile +import argparse +from json import dump, load +from zipfile import ZipFile +from tempfile import gettempdir + +RootPackURL = "http://www.keil.com/pack/index.idx" + +LocalPackDir = dirname(__file__) +LocalPackIndex = join(LocalPackDir, "index.json") +LocalPackAliases = join(LocalPackDir, "aliases.json") + + +protocol_matcher = compile("\w*://") +def strip_protocol(url) : + return protocol_matcher.sub("", str(url)) + +def largest_version(content) : + return sorted([t['version'] for t in content.package.releases('release')], reverse=True)[0] + +def do_queue(Class, function, interable) : + q = Queue() + threads = [Class(q, function) for each in range(20)] + for each in threads : + each.setDaemon(True) + each.start() + for thing in interable : + q.put(thing) + q.join() + +class Reader (Thread) : + def __init__(self, queue, func) : + Thread.__init__(self) + self.queue = queue + self.func = func + def run(self) : + while True : + url = self.queue.get() + self.func(url) + self.queue.task_done() + + +class Cache () : + """ The Cache object is the only relevant API object at the moment + + Constructing the Cache object does not imply any caching. + A user of the API must explicitly call caching functions. + + :param silent: A boolean that, when True, significantly reduces the printing of this Object + :type silent: bool + :param no_timeouts: A boolean that, when True, disables the default connection timeout and low speed timeout for downloading things. + :type no_timeouts: bool + """ + def __init__ (self, silent, no_timeouts) : + self.silent = silent + self.counter = 0 + self.total = 1 + self._index = {} + self._aliases = {} + self.urls = None + self.no_timeouts = no_timeouts + self.data_path = gettempdir() + + def display_counter (self, message) : + stdout.write("{} {}/{}\r".format(message, self.counter, self.total)) + stdout.flush() + + def cache_file (self, url) : + """Low level interface to caching a single file. + + :param url: The URL to cache. + :type url: str + :rtype: None + """ + if not self.silent : print("Caching {}...".format(url)) + dest = join(self.data_path, strip_protocol(url)) + try : + makedirs(dirname(dest)) + except OSError as exc : + if exc.errno == EEXIST : pass + else : raise + try: + with open(dest, "wb+") as fd : + fd.write(urlopen(url).read()) + except URLError as e: + stderr.write(e.reason) + self.counter += 1 + self.display_counter("Caching Files") + + def pdsc_to_pack (self, url) : + """Find the URL of the specified pack file described by a PDSC. + + The PDSC is assumed to be cached and is looked up in the cache by its URL. + + :param url: The url used to look up the PDSC. + :type url: str + :return: The url of the PACK file. + :rtype: str + """ + content = self.pdsc_from_cache(url) + new_url = content.package.url.get_text() + if not new_url.endswith("/") : + new_url = new_url + "/" + return (new_url + content.package.vendor.get_text() + "." + + content.package.find('name').get_text() + "." + + largest_version(content) + ".pack") + + def cache_pdsc_and_pack (self, url) : + self.cache_file(url) + try : + self.cache_file(self.pdsc_to_pack(url)) + except AttributeError : + stderr.write("[ ERROR ] {} does not appear to be a conforming .pdsc file\n".format(url)) + self.counter += 1 + + def get_urls(self): + """Extract the URLs of all know PDSC files. + + Will pull the index from the internet if it is not cached. + + :return: A list of all PDSC URLs + :rtype: [str] + """ + if not self.urls : + try : root_data = self.pdsc_from_cache(RootPackURL) + except IOError : root_data = self.cache_and_parse(RootPackURL) + self.urls = ["/".join([pdsc.get('url').strip("/"), + pdsc.get('name').strip("/")]) + for pdsc in root_data.find_all("pdsc")] + return self.urls + + def _extract_dict(self, device, filename, pack) : + to_ret = dict(pdsc_file=filename, pack_file=pack) + try : to_ret["memory"] = dict([(m["id"], dict(start=m["start"], + size=m["size"])) + for m in device("memory")]) + except (KeyError, TypeError, IndexError) as e : pass + try: algorithms = device("algorithm") + except: + try: algorithms = device.parent("algorithm") + except: pass + else: + if not algorithms: + try: algorithms = device.parent("algorithm") + except: pass + try : to_ret["algorithm"] = dict([(algo.get("name").replace('\\','/'), + dict(start=algo["start"], + size=algo["size"], + ramstart=algo.get("ramstart",None), + ramsize=algo.get("ramsize",None), + default=algo.get("default",1))) + for algo in algorithms]) + except (KeyError, TypeError, IndexError) as e: pass + try: to_ret["debug"] = device.parent.parent.debug["svd"] + except (KeyError, TypeError, IndexError) as e : pass + try: to_ret["debug"] = device.parent.debug["svd"] + except (KeyError, TypeError, IndexError) as e : pass + try: to_ret["debug"] = device.debug["svd"] + except (KeyError, TypeError, IndexError) as e : pass + + to_ret["compile"] = {} + try: compile_l1 = device.parent("compile") + except (KeyError, TypeError, IndexError) as e : compile_l1 = [] + try: compile_l2 = device.parent.parent("compile") + except (KeyError, TypeError, IndexError) as e : compile_l2 = [] + compile = compile_l2 + compile_l1 + for c in compile: + try: to_ret["compile"]["header"] = c["header"] + except (KeyError, TypeError, IndexError) as e : pass + try: to_ret["compile"]["define"] = c["define"] + except (KeyError, TypeError, IndexError) as e : pass + + try: to_ret["core"] = device.parent.processor['dcore'] + except (KeyError, TypeError, IndexError) as e : pass + try: to_ret["core"] = device.parent.parent.processor['dcore'] + except (KeyError, TypeError, IndexError) as e : pass + + to_ret["processor"] = {} + try: proc_l1 = device("processor") + except (KeyError, TypeError, IndexError) as e: proc_l1 = [] + try: proc_l2 = device.parent("processor") + except (KeyError, TypeError, IndexError) as e: proc_l2 = [] + try: proc_l3 = device.parent.parent("processor") + except (KeyError, TypeError, IndexError) as e: proc_l3 = [] + proc = proc_l3 + proc_l2 + proc_l1 + for p in proc: + try: to_ret["processor"]["fpu"] = p['dfpu'] + except (KeyError, TypeError, IndexError) as e: pass + try: to_ret["processor"]["endianness"] = p['dendian'] + except (KeyError, TypeError, IndexError) as e: pass + try: to_ret["processor"]["clock"] = p['dclock'] + except (KeyError, TypeError, IndexError) as e: pass + + try: to_ret["vendor"] = device.parent['dvendor'] + except (KeyError, TypeError, IndexError) as e: pass + try: to_ret["vendor"] = device.parent.parent['dvendor'] + except (KeyError, TypeError, IndexError) as e: pass + + if not to_ret["processor"]: + del to_ret["processor"] + + if not to_ret["compile"]: + del to_ret["compile"] + + to_ret['debug-interface'] = [] + + return to_ret + + def _generate_index_helper(self, d) : + try : + pack = self.pdsc_to_pack(d) + self._index.update(dict([(dev['dname'], self._extract_dict(dev, d, pack)) for dev in + (self.pdsc_from_cache(d)("device"))])) + except AttributeError as e : + stderr.write("[ ERROR ] file {}\n".format(d)) + print(e) + self.counter += 1 + self.display_counter("Generating Index") + + def _generate_aliases_helper(self, d) : + try : + mydict = [] + for dev in self.pdsc_from_cache(d)("board"): + try : + mydict.append((dev['name'], dev.mounteddevice['dname'])) + except (KeyError, TypeError, IndexError) as e: + pass + self._aliases.update(dict(mydict)) + except (AttributeError, TypeError) as e : + pass + self.counter += 1 + self.display_counter("Scanning for Aliases") + + def get_flash_algorthim_binary(self, device_name) : + """Retrieve the flash algorithm file for a particular part. + + Assumes that both the PDSC and the PACK file associated with that part are in the cache. + + :param device_name: The exact name of a device + :type device_name: str + :return: A file-like object that, when read, is the ELF file that describes the flashing algorithm + :rtype: ZipExtFile + """ + pack = self.pack_from_cache(self.index[device_name]) + return pack.open(device['algorithm']['file']) + + def get_svd_file(self, device_name) : + """Retrieve the flash algorithm file for a particular part. + + Assumes that both the PDSC and the PACK file associated with that part are in the cache. + + :param device_name: The exact name of a device + :type device_name: str + :return: A file-like object that, when read, is the ELF file that describes the flashing algorithm + :rtype: ZipExtFile + """ + pack = self.pack_from_cache(self.index[device_name]) + return pack.open(device['debug']) + + def generate_index(self) : + self._index = {} + self.counter = 0 + do_queue(Reader, self._generate_index_helper, self.get_urls()) + with open(LocalPackIndex, "wb+") as out: + self._index["version"] = "0.1.0" + dump(self._index, out) + stdout.write("\n") + + def generate_aliases(self) : + self._aliases = {} + self.counter = 0 + do_queue(Reader, self._generate_aliases_helper, self.get_urls()) + with open(LocalPackAliases, "wb+") as out: + dump(self._aliases, out) + stdout.write("\n") + + def find_device(self, match) : + choices = process.extract(match, self.index.keys(), limit=len(self.index)) + choices = sorted([(v, k) for k, v in choices], reverse=True) + if choices : choices = list(takewhile(lambda t: t[0] == choices[0][0], choices)) + return [(v, self.index[v]) for k,v in choices] + + def dump_index_to_file(self, file) : + with open(file, "wb+") as out: + dump(self.index, out) + + @property + def index(self) : + """An index of most of the important data in all cached PDSC files. + + :Example: + + >>> from ArmPackManager import Cache + >>> a = Cache() + >>> a.index["LPC1768"] + {u'algorithm': {u'RAMsize': u'0x0FE0', + u'RAMstart': u'0x10000000', + u'name': u'Flash/LPC_IAP_512.FLM', + u'size': u'0x80000', + u'start': u'0x00000000'}, + u'compile': [u'Device/Include/LPC17xx.h', u'LPC175x_6x'], + u'debug': u'SVD/LPC176x5x.svd', + u'pdsc_file': u'http://www.keil.com/pack/Keil.LPC1700_DFP.pdsc', + u'memory': {u'IRAM1': {u'size': u'0x8000', u'start': u'0x10000000'}, + u'IRAM2': {u'size': u'0x8000', u'start': u'0x2007C000'}, + u'IROM1': {u'size': u'0x80000', u'start': u'0x00000000'}}} + + + """ + if not self._index : + with open(LocalPackIndex) as i : + self._index = load(i) + return self._index + @property + def aliases(self) : + """An index of most of the important data in all cached PDSC files. + + :Example: + + >>> from ArmPackManager import Cache + >>> a = Cache() + >>> a.index["LPC1768"] + {u'algorithm': {u'RAMsize': u'0x0FE0', + u'RAMstart': u'0x10000000', + u'name': u'Flash/LPC_IAP_512.FLM', + u'size': u'0x80000', + u'start': u'0x00000000'}, + u'compile': [u'Device/Include/LPC17xx.h', u'LPC175x_6x'], + u'debug': u'SVD/LPC176x5x.svd', + u'pdsc_file': u'http://www.keil.com/pack/Keil.LPC1700_DFP.pdsc', + u'memory': {u'IRAM1': {u'size': u'0x8000', u'start': u'0x10000000'}, + u'IRAM2': {u'size': u'0x8000', u'start': u'0x2007C000'}, + u'IROM1': {u'size': u'0x80000', u'start': u'0x00000000'}}} + + + """ + if not self._aliases : + with open(join(self.data_path, "aliases.json")) as i : + self._aliases = load(i) + return self._aliases + + def cache_everything(self) : + """Cache every PACK and PDSC file known. + + Generates an index afterwards. + + .. note:: This process may use 4GB of drive space and take upwards of 10 minutes to complete. + """ + self.cache_pack_list(self.get_urls()) + self.generate_index() + self.generate_aliases() + + def cache_descriptors(self) : + """Cache every PDSC file known. + + Generates an index afterwards. + + .. note:: This process may use 11MB of drive space and take upwards of 1 minute. + """ + self.cache_descriptor_list(self.get_urls()) + self.generate_index() + self.generate_aliases() + + def cache_descriptor_list(self, list) : + """Cache a list of PDSC files. + + :param list: URLs of PDSC files to cache. + :type list: [str] + """ + self.total = len(list) + self.display_counter("Caching Files") + do_queue(Reader, self.cache_file, list) + stdout.write("\n") + + def cache_pack_list(self, list) : + """Cache a list of PACK files, referenced by their PDSC URL + + :param list: URLs of PDSC files to cache. + :type list: [str] + """ + self.total = len(list) * 2 + self.display_counter("Caching Files") + do_queue(Reader, self.cache_pdsc_and_pack, list) + stdout.write("\n") + + def pdsc_from_cache(self, url) : + """Low level inteface for extracting a PDSC file from the cache. + + Assumes that the file specified is a PDSC file and is in the cache. + + :param url: The URL of a PDSC file. + :type url: str + :return: A parsed representation of the PDSC file. + :rtype: BeautifulSoup + """ + dest = join(self.data_path, strip_protocol(url)) + with open(dest, "r") as fd : + return BeautifulSoup(fd, "html.parser") + + def pack_from_cache(self, url) : + """Low level inteface for extracting a PACK file from the cache. + + Assumes that the file specified is a PACK file and is in the cache. + + :param url: The URL of a PACK file. + :type url: str + :return: A parsed representation of the PACK file. + :rtype: ZipFile + """ + return ZipFile(join(self.data_path, + strip_protocol(device['pack_file']))) + + def gen_dict_from_cache() : + pdsc_files = pdsc_from_cache(RootPackUrl) + + def cache_and_parse(self, url) : + """A low level shortcut that Caches and Parses a PDSC file. + + :param url: The URL of the PDSC file. + :type url: str + :return: A parsed representation of the PDSC file. + :rtype: BeautifulSoup + """ + self.cache_file(url) + return self.pdsc_from_cache(url) +