Project

General

Profile

1 13 aaronmk
# XML-database conversion
2
3 2418 aaronmk
import copy
4 16 aaronmk
import re
5 13 aaronmk
from xml.dom import Node
6
7 446 aaronmk
import exc
8 2436 aaronmk
import Parser
9 46 aaronmk
import sql
10 3077 aaronmk
import sql_io
11 2268 aaronmk
import sql_gen
12 84 aaronmk
import strings
13 133 aaronmk
import util
14 46 aaronmk
import xml_dom
15 2113 aaronmk
import xml_func
16 2436 aaronmk
import xpath
17 13 aaronmk
18 139 aaronmk
def name_of(node): return re.sub(r'^.*\.', r'', node.tagName)
19 16 aaronmk
20 13 aaronmk
ptr_suffix = '_id'
21
22 138 aaronmk
def is_ptr(node_name): return node_name.lower().endswith(ptr_suffix)
23 13 aaronmk
24 172 aaronmk
def ptr_type_guess(node_name):
25 13 aaronmk
    assert is_ptr(node_name)
26
    return node_name[:-len(ptr_suffix)]
27
28
def ptr_target(node):
29
    assert is_ptr(name_of(node))
30 2113 aaronmk
    return xml_dom.value_node(node)
31 13 aaronmk
32
def find_by_name(node, name):
33 46 aaronmk
    for parent in xml_dom.NodeParentIter(node):
34 13 aaronmk
        if name_of(parent) == name: return parent
35
        else:
36 46 aaronmk
            for child in xml_dom.NodeElemIter(parent):
37 16 aaronmk
                child_name = name_of(child)
38 13 aaronmk
                if is_ptr(child_name):
39 172 aaronmk
                    target = ptr_target(child)
40
                    if target.tagName == name: return target
41 13 aaronmk
                elif child_name == name: return child
42
    return None
43
44 1850 aaronmk
def get(db, node, limit=None, start=None):
45
    def pkey(table): return sql.pkey(db, table)
46 126 aaronmk
47 141 aaronmk
    node = node.firstChild
48 135 aaronmk
    table = name_of(node)
49
    pkey_ = pkey(table)
50
51
    fields = []
52
    conds = {}
53 133 aaronmk
    for child in xml_dom.NodeElemIter(node):
54 135 aaronmk
        child_name = name_of(child)
55
        if xml_dom.is_empty(child): fields.append(child_name)
56
        elif xml_dom.is_text(child): conds[child_name] = xml_dom.value(child)
57
        else: raise Exception('Joins not supported yet')
58
    id_ = xml_dom.get_id(node)
59 1836 aaronmk
    if id_ != None: conds[pkey(table)] = id_ # replace any existing pkey value
60 135 aaronmk
    if fields == []: fields.append(pkey_)
61 133 aaronmk
62 864 aaronmk
    return sql.select(db, table, fields, conds, limit, start)
63 126 aaronmk
64 1864 aaronmk
def put(db, node, row_ct_ref=None, on_error=exc.raise_, pool=None,
65
    store_ids=False, parent_id=None):
66 138 aaronmk
    '''store_ids enables searching the tree for missing fields'''
67 1850 aaronmk
    def pkey(table): return sql.pkey(db, table, True)
68 15 aaronmk
69 461 aaronmk
    def put_(node, parent_id=None):
70 1874 aaronmk
        args = (db, node, row_ct_ref, on_error, pool, store_ids, parent_id)
71
        if parent_id != None and pool != None: pool.apply_async(put, args)
72
        else: return put(*args)
73 461 aaronmk
74
    def on_error_(e):
75 446 aaronmk
        exc.add_msg(e, 'node:\n'+str(node))
76 461 aaronmk
        on_error(e)
77 446 aaronmk
78 48 aaronmk
    table = name_of(node)
79 446 aaronmk
    try: pkey_ = pkey(table)
80 461 aaronmk
    except sql.DatabaseErrors, e: on_error_(e); return None
81 48 aaronmk
    row = {}
82
    children = []
83 13 aaronmk
84 48 aaronmk
    # Divide children into fields and children with fkeys to parent
85
    for child in xml_dom.NodeElemIter(node):
86
        child_name = name_of(child)
87 463 aaronmk
        if xml_dom.is_empty(child): row[child_name] = None
88 454 aaronmk
        elif xml_dom.is_text(child):
89 84 aaronmk
            row[child_name] = strings.to_unicode(xml_dom.value(child))
90 461 aaronmk
        elif is_ptr(child_name): row[child_name] = put_(ptr_target(child))
91 48 aaronmk
        else: children.append(child)
92
93
    # Add fkey to parent
94 59 aaronmk
    if parent_id != None:
95
        parent_ptr = node.getAttribute('fkey')
96
        if parent_ptr == '': parent_ptr = pkey(name_of(node.parentNode))
97
        row[parent_ptr] = parent_id
98 48 aaronmk
99
    # Insert node
100 446 aaronmk
    try:
101
        for try_num in xrange(2):
102
            try:
103 3077 aaronmk
                id_ = sql_io.put(db, table, row, pkey_, row_ct_ref)
104 446 aaronmk
                if store_ids: xml_dom.set_id(node, id_)
105
                break
106
            except sql.NullValueException, e:
107 468 aaronmk
                col = e.cols[0]
108 446 aaronmk
                if try_num > 0: raise # exception still raised after retry
109 468 aaronmk
                if store_ids and is_ptr(col):
110 446 aaronmk
                    # Search for required column in ancestors and their children
111 468 aaronmk
                    target = find_by_name(node, ptr_type_guess(col))
112 446 aaronmk
                    if target == None: raise
113 468 aaronmk
                    row[col] = xml_dom.get_id(target)
114 446 aaronmk
                else: raise
115 461 aaronmk
    except sql.DatabaseErrors, e: on_error_(e); return None
116 48 aaronmk
117
    # Insert children with fkeys to parent
118 1874 aaronmk
    for child in children: put_(child, id_)
119 48 aaronmk
120
    return id_
121 1996 aaronmk
122 2039 aaronmk
class ColRef:
123
    '''A reference to a table column'''
124
    def __init__(self, name, idx):
125
        self.name = name
126
        self.idx = idx
127
128
    def __str__(self): return self.name
129
130 3114 aaronmk
# Controls when and how put_table() will partition the input table
131 3145 aaronmk
partition_size = 500000 # rows
132 3114 aaronmk
133 2177 aaronmk
input_col_prefix = '$'
134
135 2549 aaronmk
put_table_special_funcs = set(['_simplifyPath'])
136
137 3622 aaronmk
def put_table(db, node, in_table, in_row_ct_ref=None, row_ins_ct_ref=None,
138
    limit=None, start=0, on_error=exc.raise_, col_defaults={}):
139 1996 aaronmk
    '''
140 1998 aaronmk
    @param node The XML tree that transforms the input to the output. Similar to
141 2177 aaronmk
        put()'s node param, but with the input column name prefixed by
142
        input_col_prefix in place of the column value.
143 3194 aaronmk
    @return sql_gen.Col Where the pkeys (from INSERT RETURNING) are made
144 2067 aaronmk
        available
145 1998 aaronmk
    '''
146 2418 aaronmk
    in_table = sql_gen.as_Table(in_table)
147 3113 aaronmk
    in_table.set_srcs([in_table], overwrite=False)
148 3184 aaronmk
    db.src = str(in_table)
149 2418 aaronmk
150 3621 aaronmk
    db.autoanalyze = True # but don't do this in row-based import
151
    db.autoexplain = True # but don't do this in row-based import
152 3107 aaronmk
153 3621 aaronmk
    full_in_table = in_table
154
155 3620 aaronmk
    # Subset and partition in_table
156
    # OK to do even if table already the right size because it takes <1 sec.
157 3621 aaronmk
    total = 0
158
    while limit == None or total < limit:
159
        # Adjust partition size if last partition
160
        this_limit = partition_size
161
        if limit != None: this_limit = min(this_limit, limit - total)
162 3268 aaronmk
163 3621 aaronmk
        # Row # is interally 0-based, but 1-based to the user
164
        db.log_debug('********** Partition: rows '+str(start+1)+'-'
165
            +str(start+this_limit)+' **********', level=1.2)
166 3122 aaronmk
167 3621 aaronmk
        # Subset in_table
168
        in_table = copy.copy(full_in_table) # don't modify input!
169
        in_table.name = str(in_table) # prepend schema
170
        cur = sql.run_query_into(db, sql.mk_select(db, full_in_table,
171
            limit=this_limit, start=start), into=in_table, add_pkey_=True)
172
            # full_in_table will be shadowed (hidden) by created temp table
173 3121 aaronmk
174 3621 aaronmk
        this_ct = cur.rowcount
175
        total += this_ct
176
        start += this_ct # advance start to fetch next set
177
        if this_ct == 0: break # in_table size is multiple of partition_size
178
179
        # Recurse
180
        pkeys_loc = _put_table_part(db, node, in_table, row_ins_ct_ref,
181 3622 aaronmk
            on_error, col_defaults)
182 3621 aaronmk
        if in_row_ct_ref != None: in_row_ct_ref[0] += this_ct
183
184
        sql.empty_temp(db, in_table)
185
186
        if this_ct < partition_size: break # partial partition = last
187
188
        # Work around PostgreSQL's temp table disk space leak
189
        db.reconnect()
190 3114 aaronmk
191 3621 aaronmk
    return pkeys_loc
192
193 3622 aaronmk
def _put_table_part(db, node, in_table, row_ins_ct_ref, on_error, col_defaults,
194
    parent_ids_loc=None, next=None):
195 3621 aaronmk
    '''Helper function for put_table() only; should not be called directly'''
196
    def put_table_(node, in_row_ct_ref=None):
197
        return _put_table_part(db, node, in_table, row_ins_ct_ref, on_error,
198 3622 aaronmk
            col_defaults, parent_ids_loc, next)
199 3621 aaronmk
200 2551 aaronmk
    is_func = xml_func.is_func(node)
201 2005 aaronmk
    out_table = name_of(node)
202 2432 aaronmk
203
    # Divide children into fields and children with fkeys to parent
204 2005 aaronmk
    row = {}
205
    children = []
206
    for child in xml_dom.NodeElemIter(node):
207
        child_name = name_of(child)
208
        if xml_dom.is_empty(child): row[child_name] = None
209
        elif xml_dom.is_text(child):
210
            row[child_name] = strings.to_unicode(xml_dom.value(child))
211 2113 aaronmk
        else:
212
            child_value = xml_dom.value_node(child)
213 2434 aaronmk
            if is_func or is_ptr(child_name) or xml_func.is_func(child_value):
214 2432 aaronmk
                row[child_name] = child_value
215 2113 aaronmk
            else: children.append(child)
216 2434 aaronmk
217
    # Special handling for structural XML functions
218
    if out_table == '_simplifyPath':
219 2436 aaronmk
        # Parse args
220
        def wrap_e(e): raise xml_func.SyntaxError(e)
221 2434 aaronmk
        try:
222 2506 aaronmk
            next = row['next'] # modifies outer next var used by put_table_()
223 2434 aaronmk
            require = row['require']
224
            path = row['path']
225 2436 aaronmk
        except KeyError, e: wrap_e(e)
226
        try: next = xpath.parse(next)
227
        except Parser.SyntaxError, e: wrap_e(e)
228
        try: next = next[0].name
229
        except IndexError, e: wrap_e(e)
230 2434 aaronmk
231 2506 aaronmk
        return put_table_(path)
232 2434 aaronmk
233 2966 aaronmk
    no_empty = set([in_table])
234
235 3107 aaronmk
    def pkey(table): return sql.pkey(db, table, True)
236
237 2177 aaronmk
    # Add fkey to parent
238
    if parent_ids_loc != None:
239 3601 aaronmk
        if sql_gen.is_table_col(parent_ids_loc):
240
            no_empty.add(parent_ids_loc.table)
241 2177 aaronmk
        parent_ptr = node.getAttribute('fkey')
242
        if parent_ptr == '': parent_ptr = pkey(name_of(node.parentNode))
243
        row[parent_ptr] = parent_ids_loc
244
245 2060 aaronmk
    # Divide fields into input columns and literal values
246 2506 aaronmk
    parent_ids_loc = None # applies to this section
247 2313 aaronmk
    in_tables = [in_table]
248 2060 aaronmk
    for out_col, value in row.iteritems():
249 2432 aaronmk
        # Handle forward pointers
250
        if xml_dom.is_node(value): row[out_col] = value = put_table_(value)
251
252
        # Translate values
253 2268 aaronmk
        if isinstance(value, sql_gen.Col): # value is temp table column
254
            in_tables.append(value.table)
255 2177 aaronmk
        elif util.is_str(value) and value.startswith(input_col_prefix):
256
            # value is input column
257 2272 aaronmk
            row[out_col] = sql_gen.Col(strings.remove_prefix(input_col_prefix,
258 2313 aaronmk
                value), in_table)
259 3428 aaronmk
        else: # value is literal value
260 2323 aaronmk
            row[out_col] = sql_gen.NamedCol(out_col, value)
261 2060 aaronmk
262
    # Insert node
263 3077 aaronmk
    pkeys_loc = sql_io.put_table(db, out_table, in_tables, row, row_ins_ct_ref,
264 3619 aaronmk
        None, next, col_defaults, on_error)
265 2060 aaronmk
266 2975 aaronmk
    sql.empty_temp(db, set(in_tables) - no_empty)
267 2966 aaronmk
268 2177 aaronmk
    # Insert children with fkeys to parent
269 2506 aaronmk
    parent_ids_loc = pkeys_loc # applies to this section
270
    for child in children: put_table_(child)
271 2177 aaronmk
272 2133 aaronmk
    return pkeys_loc