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(['_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={},
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
    
66
    def put_(node):
67
        return put(db, node, row_ins_ct_ref, on_error, col_defaults, in_table,
68
            parent_ids_loc, next)
69
    
70
    def augment_error(e): exc.add_msg(e, 'node:\n'+strings.ustr(node))
71
    def on_error_(e):
72
        augment_error(e)
73
        on_error(e)
74
    
75
    def wrap_e(e):
76
        augment_error(e)
77
        raise xml_func.SyntaxError(e)
78
    
79
    is_func = xml_func.is_func(node)
80
    out_table = name_of(node)
81
    
82
    # Divide children into fields and children with fkeys to parent
83
    row = dicts.OnceOnlyDict()
84
    children = []
85
    try:
86
        for child in xml_dom.NodeElemIter(node):
87
            child_name = name_of(child)
88
            if xml_dom.is_empty(child): row[child_name] = None
89
            elif xml_dom.is_text(child):
90
                row[child_name] = strings.to_unicode(xml_dom.value(child))
91
            else:
92
                child_value = xml_dom.value_node(child)
93
                if ((is_func or is_ptr(child_name)
94
                    or xml_func.is_func(child_value))
95
                    and not xml_func.is_func(child)):
96
                    row[child_name] = child_value
97
                else: children.append(child)
98
    except dicts.KeyExistsError, e: wrap_e(e)
99
    
100
    # Special handling for structural XML functions
101
    if out_table == '_simplifyPath':
102
        # Parse args
103
        try:
104
            next = row['next'] # modifies outer next var used by put_()
105
            path = row['path']
106
        except KeyError, e: wrap_e(e)
107
        try: next = xpath.parse(next)
108
        except Parser.SyntaxError, e: wrap_e(e)
109
        try: next = next[0].name
110
        except IndexError, e: wrap_e(e)
111
        
112
        return put_(path)
113
    
114
    is_literals = in_table == None
115
    in_tables = []
116
    no_empty = set()
117
    if not is_literals:
118
        in_tables.append(in_table)
119
        no_empty.add(in_table)
120
    
121
    def pkey_name(table): return sql.pkey_name(db, table, True)
122
    
123
    # Add fkey to parent
124
    if parent_ids_loc is not no_parent_ids_loc:
125
        if sql_gen.is_table_col(parent_ids_loc):
126
            no_empty.add(parent_ids_loc.table)
127
        parent_ptr = node.getAttribute('fkey')
128
        if parent_ptr == '': parent_ptr = pkey_name(name_of(node.parentNode))
129
        row[parent_ptr] = parent_ids_loc
130
    
131
    # Parse input columns
132
    row = row.inner # now allow keys to be overwritten
133
    for out_col, value in row.iteritems():
134
        if (not is_literals and util.is_str(value)
135
            and value.startswith(input_col_prefix)): # value is input column
136
            row[out_col] = sql_gen.Col(strings.remove_prefix(input_col_prefix,
137
                value), in_table)
138
    
139
    # Optimizations for structural XML functions
140
    if out_table == '_alt': # return first arg if non-NULL
141
        args = row.items()
142
        args.sort()
143
        out_col, value = min(args) # first arg
144
        if xml_dom.is_node(value): row[out_col] = value = put_(value)
145
        if not sql_gen.is_nullable(db, value): return value
146
    
147
    # Process values
148
    parent_ids_loc = no_parent_ids_loc # applies to this section
149
    for out_col, value in row.iteritems():
150
        # Handle forward pointers
151
        if xml_dom.is_node(value): row[out_col] = value = put_(value)
152
        
153
        # Translate values
154
        if isinstance(value, sql_gen.Col): # value is table column
155
            assert sql_gen.is_table_col(value)
156
            if value.table is not in_table: in_tables.append(value.table)
157
        else: # value is literal value
158
            row[out_col] = sql_gen.NamedCol(out_col, value)
159
    
160
    # Insert node
161
    try: pkeys_loc = sql_io.put_table(db, out_table, in_tables, row,
162
        row_ins_ct_ref, next, col_defaults, on_error_)
163
    except Exception, e:
164
        augment_error(e)
165
        raise
166
    if sql_gen.is_table_col(pkeys_loc): no_empty.add(pkeys_loc.table)
167
    
168
    sql.empty_temp(db, set(in_tables) - no_empty)
169
    
170
    # Insert children with fkeys to parent
171
    parent_ids_loc = pkeys_loc # applies to this section
172
    for child in children: put_(child)
173
    
174
    return pkeys_loc
175

    
176
def get(db, node, limit=None, start=None):
177
    def pkey_name(table): return sql.pkey_name(db, table)
178
    
179
    node = node.firstChild
180
    table = name_of(node)
181
    pkey_ = pkey_name(table)
182
    
183
    fields = []
184
    conds = {}
185
    for child in xml_dom.NodeElemIter(node):
186
        child_name = name_of(child)
187
        if xml_dom.is_empty(child): fields.append(child_name)
188
        elif xml_dom.is_text(child): conds[child_name] = xml_dom.value(child)
189
        else: raise Exception('Joins not supported yet')
190
    id_ = xml_dom.get_id(node)
191
    if id_ != None: conds[pkey_name(table)] = id_ # replace any existing value
192
    if fields == []: fields.append(pkey_)
193
    
194
    return sql.select(db, table, fields, conds, limit, start)
195

    
196
# Controls when and how put_table() will partition the input table
197
partition_size = 1000000 # rows; must be >= NCBI.nodes size
198

    
199
def put_table(db, node, in_table, in_row_ct_ref=None, row_ins_ct_ref=None,
200
    limit=None, start=0, on_error=exc.reraise, col_defaults={},
201
    partition_size=partition_size):
202
    '''
203
    @param node The XML tree that transforms the input to the output. Similar to
204
        put()'s node param, but with the input column name prefixed by
205
        input_col_prefix in place of the column value.
206
    @return sql_gen.Col Where the pkeys (from INSERT RETURNING) are made
207
        available
208
    '''
209
    in_table = sql_gen.as_Table(in_table)
210
    sql_io.mk_errors_table(db, in_table)
211
    in_table.set_srcs([in_table], overwrite=False)
212
    db.src = strings.ustr(in_table)
213
    
214
    db.autoanalyze = True # but don't do this in row-based import
215
    db.autoexplain = True # but don't do this in row-based import
216
    
217
    # Subset and partition in_table
218
    # OK to do even if table already the right size because it takes <1 sec.
219
    full_in_table = in_table
220
    pkeys_loc = None # used if loop is never executed
221
    total = 0
222
    while limit == None or total < limit:
223
        # Adjust partition size if last partition
224
        this_limit = partition_size
225
        if limit != None: this_limit = min(this_limit, limit - total)
226
        
227
        # Row # is interally 0-based, but 1-based to the user
228
        db.log_debug('********** Partition: rows '+str(start+1)+'-'
229
            +str(start+this_limit)+' **********', level=1.2)
230
        
231
        # Subset in_table
232
        in_table = sql_gen.Table(strings.ustr(full_in_table),
233
            srcs=full_in_table.srcs, is_temp=True) # prepend schema to name
234
        sql.copy_table_struct(db, full_in_table, in_table)
235
        try: sql.add_row_num(db, in_table, 'row_num')
236
        except sql.DatabaseErrors: pass # already has pkey
237
        cur = sql.insert_select(db, in_table, None, sql.mk_select(db,
238
            full_in_table, limit=this_limit, start=start))
239
        
240
        this_ct = cur.rowcount
241
        total += this_ct
242
        start += this_ct # advance start to fetch next set
243
        if this_ct == 0: break # in_table size is multiple of partition_size
244
        
245
        # Import data
246
        pkeys_loc = put(db, node, row_ins_ct_ref, on_error, col_defaults,
247
            in_table)
248
        if in_row_ct_ref != None: in_row_ct_ref[0] += this_ct
249
        
250
        sql.empty_temp(db, in_table)
251
        
252
        if this_ct < partition_size: break # partial partition = last
253
        
254
        # Work around PostgreSQL's temp table disk space leak
255
        db.reconnect()
256
    
257
    return pkeys_loc
(10-10/42)