Project

General

Profile

1
# XML-database conversion
2

    
3
import copy
4
import re
5
from xml.dom import Node
6

    
7
import dicts
8
import exc
9
import Parser
10
import sql
11
import sql_io
12
import sql_gen
13
import strings
14
import util
15
import xml_dom
16
import xml_func
17
import xpath
18

    
19
def name_of(node): return re.sub(r'^.*\.', r'', node.tagName)
20

    
21
ptr_suffix = '_id'
22

    
23
def is_ptr(node_name): return node_name.lower().endswith(ptr_suffix)
24

    
25
def ptr_type_guess(node_name):
26
    assert is_ptr(node_name)
27
    return node_name[:-len(ptr_suffix)]
28

    
29
def ptr_target(node):
30
    assert is_ptr(name_of(node))
31
    return xml_dom.value_node(node)
32

    
33
def find_by_name(node, name):
34
    for parent in xml_dom.NodeParentIter(node):
35
        if name_of(parent) == name: return parent
36
        else:
37
            for child in xml_dom.NodeElemIter(parent):
38
                child_name = name_of(child)
39
                if is_ptr(child_name):
40
                    target = ptr_target(child)
41
                    if target.tagName == name: return target
42
                elif child_name == name: return child
43
    return None
44

    
45
class ColRef:
46
    '''A reference to a table column'''
47
    def __init__(self, name, idx):
48
        self.name = name
49
        self.idx = idx
50
    
51
    def __str__(self): return self.name
52

    
53
input_col_prefix = xml_func.var_name_prefix
54

    
55
put_special_funcs = set(['_setDefault', '_simplifyPath'])
56

    
57
no_parent_ids_loc = object() # tells put() there is no parent_ids_loc
58

    
59
def put(db, node, row_ins_ct_ref=None, on_error=exc.reraise, col_defaults=None,
60
    in_table=None, parent_ids_loc=no_parent_ids_loc, next=None):
61
    '''
62
    @param node To use an entire XML document, pass root.firstChild.
63
    '''
64
    if node == None: return None # when no rows, root.firstChild == None
65
    elif xml_dom.is_text_node(node): return xml_dom.value(node)
66
    
67
    if col_defaults == None: col_defaults = {}
68
    
69
    def put_(node):
70
        if util.is_str(node): return node
71
        return put(db, node, row_ins_ct_ref, on_error, col_defaults, in_table,
72
            parent_ids_loc, next)
73
    
74
    def augment_error(e): exc.add_msg(e, 'node:\n'+strings.ustr(node))
75
    def on_error_(e):
76
        augment_error(e)
77
        on_error(e)
78
    
79
    def wrap_e(e):
80
        augment_error(e)
81
        raise xml_func.SyntaxError(e)
82
    
83
    is_func = xml_func.is_func(node)
84
    out_table = name_of(node)
85
    
86
    # Divide children into fields and children with fkeys to parent
87
    row = dicts.OnceOnlyDict()
88
    children = []
89
    try:
90
        for child in xml_dom.NodeElemIter(node):
91
            child_name = name_of(child)
92
            if xml_dom.is_empty(child): row[child_name] = None
93
            elif xml_dom.is_text(child):
94
                row[child_name] = strings.to_unicode(xml_dom.value(child))
95
            else:
96
                child_value = xml_dom.value_node(child)
97
                if ((is_func or is_ptr(child_name)
98
                    or xml_func.is_func(child_value))
99
                    and not xml_func.is_func(child)):
100
                    row[child_name] = child_value
101
                else: children.append(child)
102
    except dicts.KeyExistsError, e: wrap_e(e)
103
    
104
    # Special handling for structural XML functions
105
    if out_table == '_setDefault':
106
        # Parse args
107
        try: path = row.pop('path')
108
        except KeyError, e: wrap_e(e)
109
        
110
        col_defaults = dicts.MergeDict(dicts.WrapDict(put_, row), col_defaults)
111
        return put_(path)
112
    elif out_table == '_simplifyPath':
113
        # Parse args
114
        try:
115
            next = row['next'] # modifies outer next var used by put_()
116
            path = row['path']
117
        except KeyError, e: wrap_e(e)
118
        try: next = xpath.parse(next)
119
        except Parser.SyntaxError, e: wrap_e(e)
120
        try: next = next[0].name
121
        except IndexError, e: wrap_e(e)
122
        
123
        return put_(path)
124
    
125
    is_literals = in_table == None
126
    in_tables = []
127
    no_empty = set()
128
    if not is_literals:
129
        in_tables.append(in_table)
130
        no_empty.add(in_table)
131
    
132
    def pkey_name(table): return sql.pkey_name(db, table, True)
133
    
134
    # Add fkey to parent
135
    if parent_ids_loc is not no_parent_ids_loc:
136
        if sql_gen.is_table_col(parent_ids_loc):
137
            no_empty.add(parent_ids_loc.table)
138
        parent_ptr = node.getAttribute('fkey')
139
        if parent_ptr == '': parent_ptr = pkey_name(name_of(node.parentNode))
140
        row[parent_ptr] = parent_ids_loc
141
    
142
    # Parse input columns
143
    row = row.inner # now allow keys to be overwritten
144
    for out_col, value in row.iteritems():
145
        if (not is_literals and util.is_str(value)
146
            and value.startswith(input_col_prefix)): # value is input column
147
            row[out_col] = sql_gen.Col(strings.remove_prefix(input_col_prefix,
148
                value), in_table)
149
    
150
    # Optimizations for structural XML functions
151
    if out_table == '_alt': # return first arg if non-NULL
152
        args = row.items()
153
        args.sort()
154
        out_col, value = min(args) # first arg
155
        if xml_dom.is_node(value): row[out_col] = value = put_(value)
156
        if not sql_gen.is_nullable(db, value): return value
157
    
158
    # Process values
159
    parent_ids_loc = no_parent_ids_loc # applies to this section
160
    for out_col, value in row.iteritems():
161
        # Handle forward pointers
162
        if xml_dom.is_node(value): row[out_col] = value = put_(value)
163
        
164
        # Translate values
165
        if isinstance(value, sql_gen.Col): # value is table column
166
            assert sql_gen.is_table_col(value)
167
            if value.table is not in_table: in_tables.append(value.table)
168
        else: # value is literal value
169
            row[out_col] = sql_gen.NamedCol(out_col, value)
170
    
171
    # Insert node
172
    try: pkeys_loc = sql_io.put_table(db, out_table, in_tables, row,
173
        row_ins_ct_ref, next, col_defaults, on_error_)
174
    except Exception, e:
175
        augment_error(e)
176
        raise
177
    if sql_gen.is_table_col(pkeys_loc): no_empty.add(pkeys_loc.table)
178
    
179
    sql.empty_temp(db, set(in_tables) - no_empty)
180
    
181
    # Insert children with fkeys to parent
182
    parent_ids_loc = pkeys_loc # applies to this section
183
    for child in children: put_(child)
184
    
185
    return pkeys_loc
186

    
187
def get(db, node, limit=None, start=None):
188
    def pkey_name(table): return sql.pkey_name(db, table)
189
    
190
    node = node.firstChild
191
    table = name_of(node)
192
    pkey_ = pkey_name(table)
193
    
194
    fields = []
195
    conds = {}
196
    for child in xml_dom.NodeElemIter(node):
197
        child_name = name_of(child)
198
        if xml_dom.is_empty(child): fields.append(child_name)
199
        elif xml_dom.is_text(child): conds[child_name] = xml_dom.value(child)
200
        else: raise Exception('Joins not supported yet')
201
    id_ = xml_dom.get_id(node)
202
    if id_ != None: conds[pkey_name(table)] = id_ # replace any existing value
203
    if fields == []: fields.append(pkey_)
204
    
205
    return sql.select(db, table, fields, conds, limit, start)
206

    
207
# Controls when and how put_table() will partition the input table
208
partition_size = 1000000 # rows; must be >= NCBI.nodes size
209

    
210
def put_table(db, node, in_table, in_row_ct_ref=None, row_ins_ct_ref=None,
211
    limit=None, start=0, on_error=exc.reraise, col_defaults={},
212
    partition_size=partition_size):
213
    '''
214
    @param node The XML tree that transforms the input to the output. Similar to
215
        put()'s node param, but with the input column name prefixed by
216
        input_col_prefix in place of the column value.
217
    @return sql_gen.Col Where the pkeys (from INSERT RETURNING) are made
218
        available
219
    '''
220
    if in_table == None:
221
        return put(db, node, row_ins_ct_ref, on_error, col_defaults)
222
    
223
    in_table = sql_gen.as_Table(in_table)
224
    sql_io.mk_errors_table(db, in_table)
225
    in_table.set_srcs([in_table], overwrite=False)
226
    db.src = strings.ustr(in_table)
227
    
228
    db.autoanalyze = True # but don't do this in row-based import
229
    db.autoexplain = True # but don't do this in row-based import
230
    
231
    # Subset and partition in_table
232
    # OK to do even if table already the right size because it takes <1 sec.
233
    full_in_table = in_table
234
    pkeys_loc = None # used if loop is never executed
235
    total = 0
236
    while limit == None or total < limit:
237
        # Adjust partition size if last partition
238
        this_limit = partition_size
239
        if limit != None: this_limit = min(this_limit, limit - total)
240
        
241
        # Row # is interally 0-based, but 1-based to the user
242
        db.log_debug('********** Partition: rows '+str(start+1)+'-'
243
            +str(start+this_limit)+' **********', level=1.2)
244
        
245
        # Subset in_table
246
        in_table = sql_gen.Table(strings.ustr(full_in_table),
247
            srcs=full_in_table.srcs, is_temp=True) # prepend schema to name
248
        sql.copy_table_struct(db, full_in_table, in_table)
249
        try: sql.add_row_num(db, in_table, 'row_num')
250
        except sql.DatabaseErrors: pass # already has pkey
251
        cur = sql.insert_select(db, in_table, None, sql.mk_select(db,
252
            full_in_table, limit=this_limit, start=start))
253
        
254
        this_ct = cur.rowcount
255
        total += this_ct
256
        start += this_ct # advance start to fetch next set
257
        if this_ct == 0: break # in_table size is multiple of partition_size
258
        
259
        # Import data
260
        pkeys_loc = put(db, node, row_ins_ct_ref, on_error, col_defaults,
261
            in_table)
262
        if in_row_ct_ref != None: in_row_ct_ref[0] += this_ct
263
        
264
        sql.empty_temp(db, in_table)
265
        
266
        if this_ct < partition_size: break # partial partition = last
267
        
268
        # Work around PostgreSQL's temp table disk space leak
269
        db.reconnect()
270
    
271
    return pkeys_loc
(12-12/44)