1 """ SVN client functions """
3 from shell
import run_svn
4 from errors
import EmptySVNLog
13 from xml
.etree
import cElementTree
as ET
16 from xml
.etree
import ElementTree
as ET
19 import cElementTree
as ET
21 from elementtree
import ElementTree
as ET
23 _identity_table
= "".join(map(chr, range(256)))
24 _forbidden_xml_chars
= "".join(
25 set(map(chr, range(32))) - set('\x09\x0A\x0D')
29 def strip_forbidden_xml_chars(xml_string
):
31 Given an XML string, strips forbidden characters as per the XML spec.
32 (these are all control characters except 0x9, 0xA and 0xD).
34 return xml_string
.translate(_identity_table
, _forbidden_xml_chars
)
36 def safe_path(path
, rev_number
=None):
38 Build a path to pass as a SVN command-line arg.
40 # URL-escape URL's, but leave local WC paths alone
42 path
= urllib
.quote(path
, ":/")
44 if rev_number
is not None:
45 path
+= "@"+str(rev_number
)
46 # Else, if path already contains an "@", add a trailing "@" to "escape" the earlier "@".
51 def svn_date_to_timestamp(svn_date
):
53 Parse an SVN date as read from the XML output and return the corresponding
56 # Strip microseconds and timezone (always UTC, hopefully)
57 # XXX there are various ISO datetime parsing routines out there,
58 # cf. http://seehuhn.de/comp/pdate
59 date
= svn_date
.split('.', 2)[0]
60 time_tuple
= time
.strptime(date
, "%Y-%m-%dT%H:%M:%S")
61 return calendar
.timegm(time_tuple
)
63 def parse_svn_info_xml(xml_string
):
65 Parse the XML output from an "svn info" command and extract useful information
69 xml_string
= strip_forbidden_xml_chars(xml_string
)
70 tree
= ET
.fromstring(xml_string
)
71 entry
= tree
.find('.//entry')
72 d
['url'] = entry
.find('url').text
73 d
['kind'] = entry
.get('kind')
74 d
['revision'] = int(entry
.get('revision'))
75 d
['repos_url'] = tree
.find('.//repository/root').text
76 d
['repos_uuid'] = tree
.find('.//repository/uuid').text
77 d
['last_changed_rev'] = int(tree
.find('.//commit').get('revision'))
78 author_element
= tree
.find('.//commit/author')
79 if author_element
is not None:
80 d
['last_changed_author'] = author_element
.text
81 d
['last_changed_date'] = svn_date_to_timestamp(tree
.find('.//commit/date').text
)
84 def get_kind(svn_repos_url
, svn_path
, svn_rev
, action
, paths
):
86 Calculate the "kind"-type of a given URL in the SVN repo.
88 # By default, just do a simple "svn info" based on passed-in params.
92 # For deletions, we can't do an "svn info" at this revision.
93 # Need to trace ancestry backwards.
96 # Build a list of any copy-from's in this log_entry that we're a child of.
97 if p
['kind'] == 'dir' and p
['copyfrom_revision'] and svn_path
.startswith(p
['path']+"/"):
98 parents
.append(p
['path'])
100 # Use the nearest copy-from'd parent
102 parent
= parents
[len(parents
)-1]
104 if parent
== p
['path']:
105 info_path
= info_path
.replace(p
['path'], p
['copyfrom_path'])
106 info_rev
= p
['copyfrom_revision']
108 # If no parent copy-from's, then we should be able to check this path in
109 # the preceeding revision.
111 info
= get_svn_info(svn_repos_url
+info_path
, info_rev
)
114 def parse_svn_log_xml(xml_string
):
116 Parse the XML output from an "svn log" command and extract useful information
117 as a list of dicts (one per log changeset).
120 xml_string
= strip_forbidden_xml_chars(xml_string
)
121 tree
= ET
.fromstring(xml_string
)
122 for entry
in tree
.findall('logentry'):
124 d
['revision'] = int(entry
.get('revision'))
125 # Some revisions don't have authors, most notably the first revision
127 # logentry nodes targeting directories protected by path-based
128 # authentication have no child nodes at all. We return an entry
129 # in that case. Anyway, as it has no path entries, no further
130 # processing will be made.
131 author
= entry
.find('author')
132 date
= entry
.find('date')
133 msg
= entry
.find('msg')
134 d
['author'] = author
is not None and author
.text
or "No author"
135 d
['date_raw'] = date
.text
if date
is not None else None
136 d
['date'] = svn_date_to_timestamp(date
.text
) if date
is not None else None
137 d
['message'] = msg
is not None and msg
.text
and msg
.text
.replace('\r\n', '\n').replace('\n\r', '\n').replace('\r', '\n') or ""
139 for path
in entry
.findall('.//paths/path'):
140 copyfrom_rev
= path
.get('copyfrom-rev')
142 copyfrom_rev
= int(copyfrom_rev
)
145 'kind': path
.get('kind'),
146 'action': path
.get('action'),
147 'copyfrom_path': path
.get('copyfrom-path'),
148 'copyfrom_revision': copyfrom_rev
,
150 # Sort paths (i.e. into hierarchical order), so that process_svn_log_entry()
151 # can process actions in depth-first order.
152 d
['changed_paths'] = sorted(paths
, key
=operator
.itemgetter('path'))
154 for prop
in entry
.findall('.//revprops/property'):
155 revprops
.append({ 'name': prop.get('name'), 'value': prop.text }
)
156 d
['revprops'] = revprops
160 def parse_svn_status_xml(xml_string
, base_dir
=None, ignore_externals
=False):
162 Parse the XML output from an "svn status" command and extract useful info
163 as a list of dicts (one per status entry).
166 base_dir
= os
.path
.normcase(base_dir
)
168 xml_string
= strip_forbidden_xml_chars(xml_string
)
169 tree
= ET
.fromstring(xml_string
)
170 for entry
in tree
.findall('.//entry'):
172 path
= entry
.get('path')
173 if base_dir
is not None and os
.path
.normcase(path
).startswith(base_dir
):
174 path
= path
[len(base_dir
):].lstrip('/\\')
176 wc_status
= entry
.find('wc-status')
177 if wc_status
.get('item') == 'external':
180 status
= wc_status
.get('item')
181 revision
= wc_status
.get('revision')
182 if status
== 'external':
183 d
['type'] = 'external'
184 elif revision
is not None:
187 d
['type'] = 'unversioned'
189 d
['revision'] = revision
190 d
['props'] = wc_status
.get('props')
191 d
['copied'] = wc_status
.get('copied')
195 def get_svn_rev(svn_url_or_wc
, rev_number
):
197 Evaluate a given SVN revision pattern, to map it to a discrete rev #.
199 xml_string
= run_svn(['info', '--xml', '-r', rev_number
, safe_path(svn_url_or_wc
, rev_number
)], fail_if_stderr
=True)
200 info
= parse_svn_info_xml(xml_string
)
201 return info
['revision']
203 def get_svn_info(svn_url_or_wc
, rev_number
=None):
205 Get SVN information for the given URL or working copy, with an optionally
206 specified revision number.
207 Returns a dict as created by parse_svn_info_xml().
209 args
= ['info', '--xml']
210 if rev_number
is not None:
211 args
+= ["-r", rev_number
]
212 args
+= [safe_path(svn_url_or_wc
, rev_number
)]
213 xml_string
= run_svn(args
, fail_if_stderr
=True)
214 return parse_svn_info_xml(xml_string
)
216 def svn_checkout(svn_url
, checkout_dir
, rev_number
=None):
218 Checkout the given URL at an optional revision number.
220 args
= ['checkout', '-q']
221 if rev_number
is not None:
222 args
+= ['-r', rev_number
]
223 args
+= [safe_path(svn_url
, rev_number
), checkout_dir
]
226 def run_svn_log(svn_url_or_wc
, rev_start
, rev_end
, limit
, stop_on_copy
=False, get_changed_paths
=True, get_revprops
=False):
228 Fetch up to 'limit' SVN log entries between the given revisions.
230 args
= ['log', '--xml']
232 args
+= ['--stop-on-copy']
233 if get_changed_paths
:
236 args
+= ['--with-all-revprops']
237 args
+= ['-r', '%s:%s' % (rev_start
, rev_end
)]
238 args
+= ['--limit', str(limit
), safe_path(svn_url_or_wc
, max(rev_start
, rev_end
))]
239 xml_string
= run_svn(args
)
240 return parse_svn_log_xml(xml_string
)
242 def get_svn_status(svn_wc
, quiet
=False, no_recursive
=False):
244 Get SVN status information about the given working copy.
246 # Ensure proper stripping by canonicalizing the path
247 svn_wc
= os
.path
.abspath(svn_wc
)
248 args
= ['status', '--xml', '--ignore-externals']
255 xml_string
= run_svn(args
+ [safe_path(svn_wc
)])
256 return parse_svn_status_xml(xml_string
, svn_wc
, ignore_externals
=True)
258 def get_svn_versioned_files(svn_wc
):
260 Get the list of versioned files in the SVN working copy.
263 for e
in get_svn_status(svn_wc
):
264 if e
['path'] and e
['type'] == 'normal':
265 contents
.append(e
['path'])
268 def get_one_svn_log_entry(svn_url
, rev_start
, rev_end
, stop_on_copy
=False, get_changed_paths
=True, get_revprops
=False):
270 Get the first SVN log entry in the requested revision range.
272 entries
= run_svn_log(svn_url
, rev_start
, rev_end
, 1, stop_on_copy
, get_changed_paths
, get_revprops
)
275 raise EmptySVNLog("No SVN log for %s between revisions %s and %s" %
276 (svn_url
, rev_start
, rev_end
))
278 def get_first_svn_log_entry(svn_url
, rev_start
, rev_end
, stop_on_copy
=True, get_changed_paths
=True):
280 Get the first log entry after (or at) the given revision number in an SVN branch.
281 By default the revision number is set to 0, which will give you the log
282 entry corresponding to the branch creaction.
284 NOTE: to know whether the branch creation corresponds to an SVN import or
285 a copy from another branch, inspect elements of the 'changed_paths' entry
286 in the returned dictionary.
288 return get_one_svn_log_entry(svn_url
, rev_start
, rev_end
, stop_on_copy
=stop_on_copy
, get_changed_paths
=get_changed_paths
)
290 def get_last_svn_log_entry(svn_url
, rev_start
, rev_end
, stop_on_copy
=True, get_changed_paths
=True):
292 Get the last log entry before/at the given revision number in an SVN branch.
293 By default the revision number is set to HEAD, which will give you the log
294 entry corresponding to the latest commit in branch.
296 return get_one_svn_log_entry(svn_url
, rev_end
, rev_start
, stop_on_copy
=stop_on_copy
, get_changed_paths
=get_changed_paths
)
299 log_duration_threshold
= 10.0
300 log_min_chunk_length
= 10
301 log_max_chunk_length
= 10000
303 def iter_svn_log_entries(svn_url
, first_rev
, last_rev
, stop_on_copy
=False, get_changed_paths
=True, get_revprops
=False, ancestors
=[]):
305 Iterate over SVN log entries between first_rev and last_rev.
307 This function features chunked log fetching so that it isn't too nasty
308 to the SVN server if many entries are requested.
310 NOTE: If *not* passing in the explicit (pre-calculated) 'ancestors' list,
311 this chunked log fetching *ONLY* works correctly on paths which
312 are known to have existed unbroken in the SVN repository, e.g. /trunk.
313 Chunked fetching breaks down if a path existed in earlier, then was
314 deleted, and later was re-created. For example, if path was created in r5,
315 then deleted in r1000, and then later re-created in r5000...
316 svn log --stop-on-copy --limit 1 -r 1:50 "path/to/file"
317 --> would yield r5, i.e. the _initial_ creation
318 svn log --stop-on-copy --limit 1 -r 1:HEAD "path/to/file"
319 --> would yield r5000, i.e. the _re-creation_
320 Use run/svn2svn.py:find_svn_ancestors() to pass in the 'ancestors' array
321 so that we can correctly re-trace ancestry here.
323 info
= get_svn_info(svn_url
)
324 svn_repos_url
= info
['repos_url']
325 #print "iter_svn_log_entries: %s %s:%s" % (svn_url, first_rev, last_rev)
326 if last_rev
== "HEAD":
327 last_rev
= info
['revision']
328 if int(first_rev
) == 1:
329 start_log
= get_first_svn_log_entry(svn_url
, first_rev
, last_rev
, stop_on_copy
=stop_on_copy
, get_changed_paths
=False)
330 if start_log
['revision'] > first_rev
:
331 first_rev
= start_log
['revision']
332 #print "first_rev: %s" % first_rev
336 cur_anc_end_rev
= None
339 # Crawl ancestry, from oldest to newest
340 for idx
in range(len(ancestors
)-1, -1, -1): # [n-1,...,0]
341 #print "(pre) Match ancestors[%s]: %s" % (idx, ancestors[idx])
342 cur_url
= svn_repos_url
+ancestors
[idx
]['copyfrom_path']
344 if first_rev
< int(ancestors
[idx
]['copyfrom_rev']):
345 cur_anc_end_rev
= int(ancestors
[idx
]['copyfrom_rev'])
347 if cur_anc_end_rev
is None:
348 #print "(pre) Match ancestors[0] (final): %s" % (ancestors[0])
350 cur_url
= svn_repos_url
+ancestors
[0]['path']
351 chunk_length
= log_min_chunk_length
352 while cur_rev
<= last_rev
:
353 #print "cur_rev:%s cur_anc_end_rev:%s cur_anc_idx:%s %s" % (cur_rev, str(cur_anc_end_rev), cur_anc_idx, cur_url)
354 if cur_anc_end_rev
and cur_rev
>= cur_anc_end_rev
:
355 cur_rev
= int(ancestors
[cur_anc_idx
]['revision'])
359 #print "(loop) Match ancestors[%s]: %s" % (idx, ancestors[idx])
360 cur_url
= svn_repos_url
+ancestors
[idx
]['copyfrom_path']
361 cur_anc_end_rev
= int(ancestors
[idx
]['copyfrom_rev'])
363 #print "(loop) Match ancestors[0] (final): %s" % (ancestors[0])
364 cur_url
= svn_repos_url
+ancestors
[0]['path']
365 cur_anc_end_rev
= None
366 #print "cur_rev:%s cur_anc_end_rev:%s cur_anc_idx:%s %s" % (cur_rev, str(cur_anc_end_rev), cur_anc_idx, cur_url)
367 start_t
= time
.time()
368 stop_rev
= min(last_rev
, cur_rev
+ chunk_length
)
369 stop_rev
= min(stop_rev
, cur_anc_end_rev
) if cur_anc_end_rev
else stop_rev
370 entries
= run_svn_log(cur_url
, cur_rev
, stop_rev
, chunk_length
,
371 stop_on_copy
, get_changed_paths
, get_revprops
)
372 duration
= time
.time() - start_t
375 if e
['revision'] > last_rev
:
377 # Embed the current URL in the yielded dict, for ancestor cases where
378 # we might have followed a copy-from to some non-original URL.
381 if e
['revision'] >= last_rev
:
383 cur_rev
= int(e
['revision'])+1
385 cur_rev
= int(stop_rev
)+1
386 # Adapt chunk length based on measured request duration
387 if duration
< log_duration_threshold
:
388 chunk_length
= min(log_max_chunk_length
, int(chunk_length
* 2.0))
389 elif duration
> log_duration_threshold
* 2:
390 chunk_length
= max(log_min_chunk_length
, int(chunk_length
/ 2.0))
393 _svn_client_version
= None
395 def get_svn_client_version():
397 Returns the SVN client version as a tuple.
399 The returned tuple only contains numbers, non-digits in version string are
402 global _svn_client_version
403 if _svn_client_version
is None:
404 raw
= run_svn(['--version', '-q']).strip()
405 _svn_client_version
= tuple(map(int, [x
for x
in raw
.split('.')
407 return _svn_client_version
410 def parse_svn_propget_xml(xml_string
):
412 Parse the XML output from an "svn propget" command and extract useful
413 information as a dict.
416 xml_string
= strip_forbidden_xml_chars(xml_string
)
417 tree
= ET
.fromstring(xml_string
)
418 prop
= tree
.find('.//property')
419 d
['name'] = prop
.get('name')
420 d
['value'] = prop
is not None and prop
.text
and prop
.text
.replace('\r\n', '\n').replace('\n\r', '\n').replace('\r', '\n') or ""
423 def parse_svn_proplist_xml(xml_string
):
425 Parse the XML output from an "svn proplist" command and extract list
429 xml_string
= strip_forbidden_xml_chars(xml_string
)
430 tree
= ET
.fromstring(xml_string
)
431 for prop
in tree
.findall('.//property'):
432 l
.append(prop
.get('name'))
435 def get_prop_value(svn_url_or_wc
, prop_name
, rev_number
=None):
437 Get the value of a versioned property for the given path.
439 args
= ['propget', '--xml']
441 args
+= ['-r', rev_number
]
442 args
+= [prop_name
, safe_path(svn_url_or_wc
, rev_number
)]
443 xml_string
= run_svn(args
)
444 return parse_svn_propget_xml(xml_string
)
446 def get_all_props(svn_url_or_wc
, rev_number
=None):
448 Get the values of all versioned properties for the given path.
451 args
= ['proplist', '--xml']
453 args
+= ['-r', rev_number
]
454 args
+= [safe_path(svn_url_or_wc
, rev_number
)]
455 xml_string
= run_svn(args
)
456 props
= parse_svn_proplist_xml(xml_string
)
457 for prop_name
in props
:
458 d
= get_prop_value(svn_url_or_wc
, prop_name
, rev_number
)
459 l
[d
['name']] = d
['value']