diff --git a/app/client/components/Importer.ts b/app/client/components/Importer.ts
index 8b5957bc..895a33ad 100644
--- a/app/client/components/Importer.ts
+++ b/app/client/components/Importer.ts
@@ -353,6 +353,7 @@ export class Importer extends DisposableWithEvents {
         label: field.label(),
         colId: destTableId ? field.colId() : null, // if inserting into new table, colId isn't defined
         type: field.column().type(),
+        widgetOptions: field.column().widgetOptions(),
         formula: field.column().formula()
       })),
       sourceCols: sourceFields.map((field) => field.colId())
diff --git a/app/client/components/TypeConversion.ts b/app/client/components/TypeConversion.ts
index b025d91c..921d3a9b 100644
--- a/app/client/components/TypeConversion.ts
+++ b/app/client/components/TypeConversion.ts
@@ -105,7 +105,7 @@ export async function prepTransformColInfo(docModel: DocModel, origCol: ColumnRe
       let {dateFormat} = prevOptions;
       if (!dateFormat) {
         const colValues = tableData.getColValues(sourceCol.colId()) || [];
-        dateFormat = guessDateFormat(colValues.map(String)) || "YYYY-MM-DD";
+        dateFormat = guessDateFormat(colValues.map(String));
       }
       widgetOptions = dateTimeWidgetOptions(dateFormat, true);
       break;
diff --git a/app/common/ActiveDocAPI.ts b/app/common/ActiveDocAPI.ts
index 7c0a369b..acafc2bb 100644
--- a/app/common/ActiveDocAPI.ts
+++ b/app/common/ActiveDocAPI.ts
@@ -49,6 +49,7 @@ export interface TransformColumn {
   colId: string|null;
   type: string;
   formula: string;
+  widgetOptions: string;
 }
 
 export interface ImportResult {
diff --git a/app/common/UserAPI.ts b/app/common/UserAPI.ts
index 90d9f509..19c94c40 100644
--- a/app/common/UserAPI.ts
+++ b/app/common/UserAPI.ts
@@ -3,7 +3,7 @@ import {ApplyUAResult, QueryFilters} from 'app/common/ActiveDocAPI';
 import {BaseAPI, IOptions} from 'app/common/BaseAPI';
 import {BillingAPI, BillingAPIImpl} from 'app/common/BillingAPI';
 import {BrowserSettings} from 'app/common/BrowserSettings';
-import {BulkColValues, TableColValues, UserAction} from 'app/common/DocActions';
+import {BulkColValues, TableColValues, TableRecordValue, TableRecordValues, UserAction} from 'app/common/DocActions';
 import {DocCreationInfo, OpenDocMode} from 'app/common/DocListAPI';
 import {Features} from 'app/common/Features';
 import {ICustomWidget} from 'app/common/CustomWidget';
@@ -402,6 +402,11 @@ export interface UserAPI {
   filters?: string;
 }
 
+interface GetRowsParams {
+  filters?: QueryFilters;
+  immediate?: boolean;
+}
+
 /**
  * Collect endpoints related to the content of a single document that we've been thinking
  * of as the (restful) "Doc API".  A few endpoints that could be here are not, for historical
@@ -411,8 +416,8 @@ export interface DocAPI {
   // Immediate flag is a currently not-advertised feature, allowing a query to proceed without
   // waiting for a document to be initialized. This is useful if the calculations done when
   // opening a document are irrelevant.
-  getRows(tableId: string, options?: { filters?: QueryFilters,
-                                       immediate?: boolean }): Promise<TableColValues>;
+  getRows(tableId: string, options?: GetRowsParams): Promise<TableColValues>;
+  getRecords(tableId: string, options?: GetRowsParams): Promise<TableRecordValue[]>;
   updateRows(tableId: string, changes: TableColValues): Promise<number[]>;
   addRows(tableId: string, additions: BulkColValues): Promise<number[]>;
   removeRows(tableId: string, removals: number[]): Promise<number[]>;
@@ -869,16 +874,13 @@ export class DocAPIImpl extends BaseAPI implements DocAPI {
     this._url = `${url}/api/docs/${docId}`;
   }
 
-  public async getRows(tableId: string, options?: { filters?: QueryFilters,
-                                                    immediate?: boolean }): Promise<TableColValues> {
-    const url = new URL(`${this._url}/tables/${tableId}/data`);
-    if (options?.filters) {
-      url.searchParams.append('filter', JSON.stringify(options.filters));
-    }
-    if (options?.immediate) {
-      url.searchParams.append('immediate', 'true');
-    }
-    return this.requestJson(url.href);
+  public async getRows(tableId: string, options?: GetRowsParams): Promise<TableColValues> {
+    return this._getRecords(tableId, 'data', options);
+  }
+
+  public async getRecords(tableId: string, options?: GetRowsParams): Promise<TableRecordValue[]> {
+    const response: TableRecordValues = await this._getRecords(tableId, 'records', options);
+    return response.records;
   }
 
   public async updateRows(tableId: string, changes: TableColValues): Promise<number[]> {
@@ -967,6 +969,17 @@ export class DocAPIImpl extends BaseAPI implements DocAPI {
     url.searchParams.append('code', code);
     return this.requestJson(url.href);
   }
+
+  private _getRecords(tableId: string, endpoint: 'data' | 'records', options?: GetRowsParams): Promise<any> {
+    const url = new URL(`${this._url}/tables/${tableId}/${endpoint}`);
+    if (options?.filters) {
+      url.searchParams.append('filter', JSON.stringify(options.filters));
+    }
+    if (options?.immediate) {
+      url.searchParams.append('immediate', 'true');
+    }
+    return this.requestJson(url.href);
+  }
 }
 
 /**
diff --git a/app/common/ValueGuesser.ts b/app/common/ValueGuesser.ts
index a3b16b1d..3c4e3050 100644
--- a/app/common/ValueGuesser.ts
+++ b/app/common/ValueGuesser.ts
@@ -162,7 +162,7 @@ export function guessColInfo(
       NumberParse.fromSettings(docSettings).guessOptions(values)
     )
       .guess(values, docSettings) ||
-    new DateGuesser(guessDateFormat(values, timezone) || "YYYY-MM-DD", timezone)
+    new DateGuesser(guessDateFormat(values, timezone), timezone)
       .guess(values, docSettings) ||
     // Don't return the same values back if there's no conversion to be done,
     // as they have to be serialized and transferred over a pipe to Python.
diff --git a/app/common/ValueParser.ts b/app/common/ValueParser.ts
index 519b1fd0..df9b9535 100644
--- a/app/common/ValueParser.ts
+++ b/app/common/ValueParser.ts
@@ -36,18 +36,6 @@ export class ValueParser {
 class IdentityParser extends ValueParser {
 }
 
-/**
- * Same as basic Value parser, but will return null if a value is an empty string.
- */
-class NullIfEmptyParser extends ValueParser {
-  public cleanParse(value: string): any {
-    if (value === "") {
-      return null;
-    }
-    return super.cleanParse(value);
-  }
-}
-
 export class NumericParser extends ValueParser {
   private _parse: NumberParse;
 
@@ -225,7 +213,6 @@ export class ReferenceListParser extends ReferenceParser {
 }
 
 export const valueParserClasses: { [type: string]: typeof ValueParser } = {
-  Any: NullIfEmptyParser,
   Numeric: NumericParser,
   Int: NumericParser,
   Date: DateParser,
diff --git a/app/common/parseDate.ts b/app/common/parseDate.ts
index a53eb1cf..0d7d98db 100644
--- a/app/common/parseDate.ts
+++ b/app/common/parseDate.ts
@@ -1,4 +1,5 @@
 import escapeRegExp = require('lodash/escapeRegExp');
+import last = require('lodash/last');
 import memoize = require('lodash/memoize');
 import {getDistinctValues, isObject} from 'app/common/gutil';
 // Simply importing 'moment-guess' inconsistently imports bundle.js or bundle.esm.js depending on environment
@@ -325,7 +326,26 @@ function standardizeTime(timeString: string): { remaining: string, time: string
   return {remaining: timeString.slice(0, match.index).trim(), time: `${hh}:${mm}:${ss}`};
 }
 
-export function guessDateFormat(values: Array<string | null>, timezone: string = 'UTC'): string | null {
+/**
+ * Guesses a full date[time] format that best matches the given strings.
+ * If several formats match equally well, picks the last one lexicographically to match the old date guessing.
+ * This means formats with an early Y and/or M are favoured.
+ * If no formats match, returns the default YYYY-MM-DD.
+ */
+export function guessDateFormat(values: Array<string | null>, timezone: string = 'UTC'): string {
+  const formats = guessDateFormats(values, timezone);
+  if (!formats) {
+    return "YYYY-MM-DD";
+  }
+  return last(formats)!;
+}
+
+/**
+ * Returns all full date[time] formats that best match the given strings.
+ * If several formats match equally well, returns them all.
+ * May return null if there are no matching formats or choosing one is too expensive.
+ */
+export function guessDateFormats(values: Array<string | null>, timezone: string = 'UTC'): string[] | null {
   const dateStrings: string[] = values.filter(isObject);
   const sample = getDistinctValues(dateStrings, 100);
   const formats: Record<string, number> = {};
@@ -358,7 +378,9 @@ export function guessDateFormat(values: Array<string | null>, timezone: string =
   }
 
   const maxCount = Math.max(...Object.values(formats));
-  return formatKeys.find(format => formats[format] === maxCount)!;
+  // Return all formats that tied for first place.
+  // Sort lexicographically for consistency in tests and with the old dateguess.py.
+  return formatKeys.filter(format => formats[format] === maxCount).sort();
 }
 
 export const dateFormatOptions = [
diff --git a/app/server/lib/ActiveDocImport.ts b/app/server/lib/ActiveDocImport.ts
index a645f95e..93e253dd 100644
--- a/app/server/lib/ActiveDocImport.ts
+++ b/app/server/lib/ActiveDocImport.ts
@@ -294,7 +294,7 @@ export class ActiveDocImport {
       const origTableName = table.table_name ? table.table_name : '';
       const transformRule = transformRuleMap && transformRuleMap.hasOwnProperty(origTableName) ?
         transformRuleMap[origTableName] : null;
-      const columnMetadata = addLabelsIfPossible(table.column_metadata);
+      const columnMetadata = cleanColumnMetadata(table.column_metadata);
       const result: ApplyUAResult = await this._activeDoc.applyUserActions(docSession,
         [["AddTable", hiddenTableName, columnMetadata]]);
       const retValue: AddTableRetValue = result.retValues[0];
@@ -313,7 +313,9 @@ export class ActiveDocImport {
       const ruleCanBeApplied = (transformRule != null) &&
                                _.difference(transformRule.sourceCols, hiddenTableColIds).length === 0;
       await this._activeDoc.applyUserActions(docSession,
-        [["ReplaceTableData", hiddenTableId, rowIdColumn, columnValues]], {parseStrings: true});
+        // BulkAddRecord rather than ReplaceTableData so that type guessing is applied to Any columns.
+        // Don't use parseStrings, only use the strict parsing in ValueGuesser to make the import lossless.
+        [["BulkAddRecord", hiddenTableId, rowIdColumn, columnValues]]);
 
       // data parsed and put into hiddenTableId
       // For preview_table (isHidden) do GenImporterView to make views and formulas and cols
@@ -433,14 +435,15 @@ export class ActiveDocImport {
 
     // If destination is a new table, we need to create it.
     if (intoNewTable) {
-      const colSpecs = destCols.map(({type, colId: id, label}) => ({type, id, label}));
+      const colSpecs = destCols.map(({type, colId: id, label, widgetOptions}) => ({type, id, label, widgetOptions}));
       const newTable = await this._activeDoc.applyUserActions(docSession, [['AddTable', destTableId, colSpecs]]);
       destTableId = newTable.retValues[0].table_id;
     }
 
     await this._activeDoc.applyUserActions(docSession,
       [['BulkAddRecord', destTableId, gutil.arrayRepeat(hiddenTableData.id.length, null), columnData]],
-      {parseStrings: true});
+      // Don't use parseStrings for new tables to make the import lossless.
+      {parseStrings: !intoNewTable});
 
     return destTableId;
   }
@@ -586,6 +589,7 @@ export class ActiveDocImport {
         colId: destTableId ? id as string : null,
         label: fields.label as string,
         type: fields.type as string,
+        widgetOptions: fields.widgetOptions as string,
         formula: srcColIds.includes(id as string) ? `$${id}` :  ''
       });
     }
@@ -730,10 +734,21 @@ function getMergeFunction({type}: MergeStrategy): MergeFunction {
 }
 
 /**
+ * Tweak the column metadata used in the AddTable action.
  * If `columns` is populated with non-blank column ids, adds labels to all
- * columns using the values set for the column ids. Otherwise, returns
- * a copy of columns with no modifications made.
+ * columns using the values set for the column ids.
+ * Ensure that columns of type Any start out as formula columns, i.e. empty columns,
+ * so that type guessing is triggered when new data is added.
  */
-function addLabelsIfPossible(columns: GristColumn[]) {
-  return columns.map(c => (c.id ? {...c, label: c.id} : c));
+function cleanColumnMetadata(columns: GristColumn[]) {
+  return columns.map(c => {
+    const newCol: any = {...c};
+    if (c.id) {
+      newCol.label = c.id;
+    }
+    if (c.type === "Any") {
+      newCol.isFormula = true;
+    }
+    return newCol;
+  });
 }
diff --git a/sandbox/grist/import_actions.py b/sandbox/grist/import_actions.py
index b65a1d37..753f7bde 100644
--- a/sandbox/grist/import_actions.py
+++ b/sandbox/grist/import_actions.py
@@ -1,12 +1,11 @@
-from collections import defaultdict, namedtuple
+from collections import namedtuple
 
-import six
-from six.moves import zip, xrange
+from six.moves import zip
 
 import column
 import identifiers
-
 import logger
+
 log = logger.Logger(__name__, logger.INFO)
 
 # Prefix for transform columns created during imports.
@@ -103,6 +102,7 @@ class ImportActions(object):
           "label":    c.label,
           "colId":    c.colId if dest_table_id else None, #should be None if into new table
           "type":     c.type,
+          "widgetOptions": getattr(c, "widgetOptions", ""),
           "formula":  ("$" + c.colId) if (c.colId in src_cols) else ''
         })
 
@@ -162,6 +162,7 @@ class ImportActions(object):
         new_col_spec = {
           "label": c.label,
           "type": c.type,
+          "widgetOptions": getattr(c, "widgetOptions", ""),
           "isFormula": True,
           "formula": c.formula}
         result = self._useractions.doAddColumn(hidden_table_id, new_col_id, new_col_spec)
diff --git a/sandbox/grist/imports/dateguess.py b/sandbox/grist/imports/dateguess.py
deleted file mode 100644
index d06bafe5..00000000
--- a/sandbox/grist/imports/dateguess.py
+++ /dev/null
@@ -1,490 +0,0 @@
-"""This module guesses possible formats of dates which can be parsed using datetime.strptime
-based on samples.
-
-dateguesser.guess(sample)
-dateguesser.guess takes a sample date string and returns a set of
-datetime.strftime/strptime-compliant date format strings that will correctly parse.
-
-dateguesser.guess_bulk(list_of_samples, error_rate=0)
-dateguesser.guess_bulk takes a list of sample date strings and acceptable error rate
-and returns a list of datetime.strftime/strptime-compliant date format strings
-sorted by error rate that will correctly parse.
-
-Algorithm:
-
-  1. Tokenize input string into chunks based on character type: digits, alphas, the rest.
-  2. Analyze each token independently in terms what format codes could represent
-  3. For given list of tokens generate all permutations of format codes
-  4. During generating permutations check for validness of generated format and skip if invalid.
-  5. Use rules listed below to decide if format is invalid:
-
-Invalid format checks:
-
-  Rule #1: Year MUST be in the date. Year is the minimum possible parsable date.
-  Rule #2. No holes (missing parts) in the format parts.
-  Rule #3. Time parts are neighbors to each other. No interleaving time with the date.
-  Rule #4. It's highly impossible that minutes coming before hour, millis coming before seconds etc
-  Rule #5. Pattern can't have some part of date/time defined more than once.
-  Rule #6: Separators between elements of the time group should be the same.
-  Rule #7: If am/pm is in date we assume that 12-hour dates are allowed only. Otherwise it's 24-hour
-  Rule #8: Year can't be between other date elements
-
-Note:
-  dateguess doesn't support defaulting to current year because parsing should be deterministic,
-  it's better to to fail guessing the format then to guess it incorrectly.
-
-Examples:
-  >>> guess('2014/05/05 14:00:00 UTC')
-  set(['%Y/%d/%m %H:%M:%S %Z', '%Y/%m/%d %H:%M:%S %Z'])
-  >>> guess('12/12/12')
-  set(['%y/%m/%d', '%d/%m/%y', '%m/%d/%y', '%y/%d/%m'])
-  >>> guess_bulk(['12-11-2014', '12-25-2014'])
-  ['%m-%d-%Y']
-  >>> guess_bulk(['12-11-2014', '25-25-2014'])
-  []
-  >>> guess_bulk(['12-11-2013', '13-8-2013', '05-25-2013', '12-25-2013'], error_rate=0.5)
-  ['%m-%d-%Y']
-"""
-
-
-import calendar
-import itertools
-import logging
-import re
-from collections import defaultdict
-
-from backports.functools_lru_cache import lru_cache
-import moment
-
-
-MONTH_NAME = calendar.month_name
-MONTH_ABBR = calendar.month_abbr
-TZ_VALID_NAMES = {z[0] for z in moment.get_tz_data().items()}
-AM_PM = {'am', 'pm'}
-DAYS_OF_WEEK_NAME = calendar.day_name
-DAYS_OF_WEEK_ABBR = calendar.day_abbr
-ASCII_DIGITS_RE = re.compile(r'^[0-9]+$')
-
-# Using x.isdigit() matches strings like u'\xb2' (superscripts) which we don't want.
-# Use isdigit(x) instead, to only match ASCII digits 0-9.
-isdigit = ASCII_DIGITS_RE.match
-
-DATE_ELEMENTS = [
-  # Name   Pattern  Predicate               Group (mutual exclusive)  Consumes N prev elements
-  ("Year", "%Y", lambda x, p, v: isdigit(x) and len(x) == 4, "Y", 0),
-  ("Year short", "%y", lambda x, p, v: isdigit(x) and len(x) == 2, "Y", 0),
-  ("Month", "%m", lambda x, p, v: isdigit(x) and len(x) <= 2 and 0 < int(x) <= 12, "m", 0),
-  ("Month name full", "%B", lambda x, p, v: x.isalpha() and x.capitalize() in MONTH_NAME, "m", 0),
-  ("Month name abbr", "%b", lambda x, p, v: x.isalpha() and x.capitalize() in MONTH_ABBR, "m", 0),
-  ("Day", "%d", lambda x, p, v: isdigit(x) and len(x) <= 2 and 0 < int(x) <= 31, "d", 0),
-  ("Day of week", "%A", lambda x, p, v: x.isalpha()
-                                        and x.capitalize() in DAYS_OF_WEEK_NAME, "a", 0),
-  ("Day of week abbr", "%a", lambda x, p, v: x.isalpha()
-                                             and x.capitalize() in DAYS_OF_WEEK_ABBR, "a", 0),
-
-  ("Compound HHMMSS", "%H%M%S", lambda x, p, v: isdigit(x) and len(x) == 6
-                                                and 0 <= int(x[0:2]) < 24
-                                                and 0 <= int(x[2:4]) < 60
-                                                and 0 <= int(x[4:6]) < 60, "HMS", 0),
-
-  ("Hour", "%H", lambda x, p, v: isdigit(x) and len(x) <= 2 and 0 <= int(x) <= 23, "H", 0),
-  ("Hour in 12hr mode", "%I", lambda x, p, v: isdigit(x) and len(x) <= 2
-                                              and 0 <= int(x) <= 11, "H", 0),
-  ("AM/PM", "%p", lambda x, p, v: x.isalpha() and len(x) == 2 and x.lower() in AM_PM, "p", 0),
-  ("Minutes", "%M", lambda x, p, v: isdigit(x) and len(x) <= 2 and 0 <= int(x) <= 59, "M", 0),
-  ("Seconds", "%S", lambda x, p, v: isdigit(x) and len(x) <= 2 and 0 <= int(x) <= 59, "S", 0),
-  ("Fraction of second", "%f", lambda x, p, v: isdigit(x) and p is not None
-                                               and p.val == '.', "f", 0),
-  ("Timezone name", "%Z", lambda x, p, v: x.isalpha() and len(x) > 2
-                                          and x in TZ_VALID_NAMES, "Z", 0),
-  ("Timezone +HHMM", "%z", lambda x, p, v: isdigit(x) and len(x) == 4 and 0 <= int(x[0:2]) < 15
-                                           and 0 <= int(x[2:4]) < 60 and p is not None
-                                           and p.val == '+', "Z", 1),
-  ("Timezone -HHMM", "%z", lambda x, p, v: isdigit(x) and len(x) == 4 and 0 <= int(x[0:2]) < 15
-                                           and 0 <= int(x[2:4]) < 60 and p is not None
-                                           and p.val == '-', "Z", 1),
-]
-
-
-class Token(object):
-  """Represents a part of a date string that's being parsed.
-  Note that __hash__ and __eq__ are overridden in order
-  to compare only meaningful parts of an object.
-  """
-  def __init__(self, val, length):
-    self.val = val
-    self.length = length
-    self.compatible_types = ()
-
-  def __hash__(self):
-    h = hash(self.length) + hash(self.compatible_types)
-    if not self.compatible_types:
-      h += hash(self.val)
-    return hash(h)
-
-  def __eq__(self, other):
-    """
-    Two tokens are equal when these both are true:
-    a) length and compatible types are equal
-    b) if it is separator (no compatible types), separator values must be equal
-    """
-    if self.length != other.length or self.compatible_types != other.compatible_types:
-      return False
-    if not other.compatible_types and self.val != other.val:
-      return False
-    return True
-
-
-def _check_rule_1(pattern, types_used):
-  """Rule #1: Year MUST be in the date. Year is the minimum possible parsable date.
-
-  Examples:
-    >>> _check_rule_1('%Y/%m/%d', 'Ymd')
-    True
-    >>> _check_rule_1('%m/%d', 'md')
-    False
-  """
-  if 'Y' not in types_used:
-    logging.debug("Rule #1 is violated for pattern %s. Types used: %s", pattern, types_used)
-    return False
-  return True
-
-
-def _check_rule_2(pattern, types_used):
-  """Rule #2: No holes (missing parts) in the format parts.
-
-  Examples:
-    >>> _check_rule_2('%Y:%H', 'YH')
-    False
-    >>> _check_rule_2('%Y/%m/%d %H', 'YmdH')
-    True
-  """
-  priorities = 'YmdHMSf'
-  seen_parts = [p in types_used for p in priorities]
-  if sorted(seen_parts, reverse=True) != seen_parts:
-    logging.debug("Rule #2 is violated for pattern %s. Types used: %s", pattern, types_used)
-    return False
-  return True
-
-
-def _check_rule_3(pattern, types_used):
-  """Rule #3: Time parts are neighbors to time only. No interleaving time with the date.
-
-  Examples:
-    >>> _check_rule_3('%m/%d %H:%M %Y', 'mdHMY')
-    True
-    >>> _check_rule_3('%m/%d %H:%Y:%M', 'mdHYM')
-    False
-  """
-  time_parts = 'HMSf'
-  time_parts_highlighted = [t in time_parts for t in types_used]
-  time_parts_deduplicated = [a[0] for a in itertools.groupby(time_parts_highlighted)]
-  if len(list(filter(lambda x: x, time_parts_deduplicated))) > 1:
-    logging.debug("Rule #3 is violated for pattern %s. Types used: %s", pattern, types_used)
-    return False
-  return True
-
-
-def _check_rule_4(pattern, types_used):
-  """Rule #4: It's highly impossible that minutes coming before hours,
-  millis coming before seconds etc.
-
-  Examples:
-    >>> _check_rule_4('%H:%M', 'HM')
-    True
-    >>> _check_rule_4('%S:%M', 'SM')
-    False
-  """
-  time_parts_priority = 'HMSf'
-  time_parts_indexes = list(filter(lambda x: x >= 0,
-                                              [time_parts_priority.find(t) for t in types_used]))
-  if sorted(time_parts_indexes) != time_parts_indexes:
-    logging.debug("Rule #4 is violated for pattern %s. Types used: %s", pattern, types_used)
-    return False
-  return True
-
-
-def _check_rule_5(pattern, types_used):
-  """Rule #5: Pattern can't have some part of date/time defined more than once.
-
-  Examples:
-    >>> _check_rule_5('%Y/%Y', 'YY')
-    False
-    >>> _check_rule_5('%m/%b', 'mm')
-    False
-    >>> _check_rule_5('%Y/%m', 'Ym')
-    True
-  """
-  if len(types_used) != len(set(types_used)):
-    logging.debug("Rule #5 is violated for pattern %s. Types used: %s", pattern, types_used)
-    return False
-  return True
-
-
-def _check_rule_6(tokens_chosen, pattern, types_used):
-  """Rule #6: Separators between elements of the time group should be the same.
-
-  Examples:
-    _check_rule_5(tokens_chosen_1, '%Y-%m-%dT%H:%M:%S', 'YmdHMS') => True
-    _check_rule_5(tokens_chosen_2, '%Y-%m-%dT%H %M %S', 'YmdHMS') => True
-    _check_rule_5(tokens_chosen_3, '%Y-%m-%dT%H-%M:%S', 'YmdHMS') => False (different separators
-                                                                  ('-' and ':') in time group)
-  """
-  time_parts = 'HMS'
-  num_of_time_parts_used = len(list(filter(lambda x: x in time_parts, types_used)))
-  time_parts_seen = 0
-  separators_seen = []
-  previous_was_a_separator = False
-
-  for token in tokens_chosen:
-    if token[1] is not None and token[1][3] in time_parts:
-      # This rule doesn't work for separator-less time group so when we found the type
-      # and it's three letters then it's (see type "Compound HHMMSS") then stop iterating
-      if len(token[1][3]) == 3:
-        break
-      # If not a first time then
-      if time_parts_seen > 0 and not previous_was_a_separator:
-        separators_seen.append(None)
-      time_parts_seen += 1
-      if time_parts_seen == num_of_time_parts_used:
-        break
-      previous_was_a_separator = False
-    else:
-      if time_parts_seen > 0:
-        separators_seen.append(token[0].val)
-      previous_was_a_separator = True
-
-  if len(set(separators_seen)) > 1:
-    logging.debug("Rule #6 is violated for pattern %s. Seen separators: %s",
-                  pattern, separators_seen)
-    return False
-  return True
-
-
-def _check_rule_7a(pattern):
-  """Rule #7a: If am/pm is in date we assume that 12-hour dates are allowed only.
-  Otherwise it's 24-hour.
-
-  Examples:
-    >>> _check_rule_7a('%Y/%m/%d %H:%M %p')
-    False
-    >>> _check_rule_7a('%Y/%m/%d %I:%M %p')
-    True
-  """
-  if '%p' in pattern and '%H' in pattern:
-    logging.debug("Rule #7a is violated for pattern %s", pattern)
-    return False
-  return True
-
-
-def _check_rule_7b(pattern):
-  """Rule #7b: If am/pm is in date we assume that 12-hour dates are allowed only.
-  Otherwise it's 24-hour.
-
-  Examples:
-    >>> _check_rule_7b('%Y/%m/%d %I:%M')
-    False
-    >>> _check_rule_7b('%Y/%m/%d %I:%M %p')
-    True
-  """
-  if '%I' in pattern and '%p' not in pattern:
-    logging.debug("Rule #7b is violated for pattern %s", pattern)
-    return False
-  return True
-
-
-def _check_rule_8(pattern, types_used):
-  """Rule #9: Year can't be between other date elements
-
-  Examples:
-    >>> _check_rule_8('%m/%Y/%d %I:%M', 'mYdIM')
-    False
-  """
-  if 'mYd' in types_used or 'dYm' in types_used:
-    logging.debug("Rule #8 is violated for pattern %s", pattern)
-    return False
-  return True
-
-
-def _tokenize_by_character_class(s):
-  """Return a list of strings by splitting s (tokenizing) by character class.
-
-  Example:
-    >>> t = _tokenize_by_character_class('Thu, May 14th, 2014 1:15 pm +0000')
-    >>> [i.val for i in t]
-    ['Thu', ',', ' ', 'May', ' ', '14', 'th', ',', ' ', '2014', ' ', '1', ':', '15', ' ', 'pm', ' ', '+', '0000']
-
-    >>> t = _tokenize_by_character_class('5/14/2014')
-    >>> [i.val for i in t]
-    ['5', '/', '14', '/', '2014']
-  """
-  res = re.split(r'(\d+)|(\W)|(_)', s)
-  return [Token(i, len(i)) for i in res if i]
-
-
-def _sliding_triplets(tokens):
-  for idx, t in enumerate(tokens):
-    yield (t, tokens[idx-1] if idx > 0 else None, tokens[idx+1] if idx < len(tokens)-1 else None)
-
-
-def _analyze_tokens(tokens):
-  """Analyze each token and find out compatible types for it."""
-  for token, prev, nxt in _sliding_triplets(tokens):
-    token.compatible_types = tuple([t for t in DATE_ELEMENTS if t[2](token.val, prev, nxt)])
-
-
-@lru_cache()
-def _generate_all_permutations(tokens):
-  """Generate all permutations of format codes for given list of tokens.
-
-  Brute-forcing of all possible permutations and rules checking eats most of the time or date
-  parsing. But since the input is expected to be highly uniform then we can expect that
-  memoization of this step will be very efficient.
-
-  Token contains values for date parts but due to overridden eq and hash methods,
-  we treat two tokens having the same length and same possible formats as equal
-  tokens and separators should be the same
-  """
-  all_patterns = set()
-  _generate_all_permutations_recursive(tokens, 0, [], "", all_patterns, "")
-
-  return all_patterns
-
-
-def _check_is_pattern_valid_quick_fail_rules(pattern, types_used):
-  """Apply rules which are applicable for partially constructed patterns.
-
-  Example: duplicates of a date part in a pattern.
-  """
-  return _check_rule_5(pattern, types_used) \
-      and _check_rule_4(pattern, types_used) \
-      and _check_rule_7a(pattern)
-
-
-def _check_is_pattern_valid_full_pattern_rules(tokens_chosen, pattern, types_used):
-  """Apply rules which are applicable for full pattern only.
-
-  Example: existence of Year part in the pattern.
-  """
-  return _check_rule_1(pattern, types_used) \
-      and _check_rule_2(pattern, types_used) \
-      and _check_rule_3(pattern, types_used) \
-      and _check_rule_6(tokens_chosen, pattern, types_used) \
-      and _check_rule_7b(pattern) \
-      and _check_rule_8(pattern, types_used)
-
-
-def _generate_all_permutations_recursive(tokens, token_idx, tokens_chosen, pattern, found_patterns,
-                                         types_used):
-  """Generate all format elements permutations recursively.
-
-  Args:
-    tokens (list[Token]): List of tokens.
-    token_idx (int): Index of token processing this cycle.
-    tokens_chosen (list[(Token, Token.compatible_type)]): List of tuples
-      containing token and compatible type
-    pattern (str): String containing format for parsing
-    found_patterns (set): Set of guessed patterns
-    types_used (str): String of types used to build pattern.
-
-  Returns:
-    list: List of permutations
-  """
-  if not _check_is_pattern_valid_quick_fail_rules(pattern, types_used):
-    return
-
-  if token_idx < len(tokens):
-    t = tokens[token_idx]
-    if t.compatible_types:
-      for ct in t.compatible_types:
-        _generate_all_permutations_recursive(tokens, token_idx+1, tokens_chosen[:] + [(t, ct)],
-                                             (pattern if ct[4] == 0 else pattern[:-ct[4]]) + ct[1],
-                                             found_patterns, types_used + ct[3])
-    else:
-      # if no compatible types it should be separator, add it to the pattern
-      _generate_all_permutations_recursive(tokens, token_idx+1,
-                                           tokens_chosen[:] + [(t, None)], pattern + t.val,
-                                           found_patterns, types_used)
-  else:
-    if _check_is_pattern_valid_full_pattern_rules(tokens_chosen, pattern, types_used):
-      found_patterns.add(pattern)
-
-
-def guess(date):
-  """Guesses datetime.strftime/strptime-compliant date formats for date string.
-
-  Args:
-    date (str): Date string.
-
-  Returns:
-    set: Set of datetime.strftime/strptime-compliant date format strings
-
-  Examples:
-    >>> guess('2014/05/05 14:00:00 UTC')
-    set(['%Y/%d/%m %H:%M:%S %Z', '%Y/%m/%d %H:%M:%S %Z'])
-    >>> guess('12/12/12')
-    set(['%y/%m/%d', '%d/%m/%y', '%m/%d/%y', '%y/%d/%m'])
-  """
-  # Don't attempt to parse strings that are so long as to be certainly non-dates. Somewhat long
-  # strings could be dates (like "Wednesday, September 16, 2020 A.D. 08:47:02.2667911 AM -06:00",
-  # and who knows what other languages do). A limit is important also because the current approach
-  # can run into "maximum recursion depth exceeded" on a very long string.
-  if len(date) > 150:
-    return set()
-  tokens = _tokenize_by_character_class(date)
-  _analyze_tokens(tokens)
-  return _generate_all_permutations(tuple(tokens))
-
-
-def guess_bulk(dates, error_rate=0):
-  """Guesses datetime.strftime/strptime-compliant date formats for list of the samples.
-
-  Args:
-    dates (list): List of samples date strings.
-    error_rate (float): Acceptable error rate (default 0.0)
-
-  Returns:
-    list: List of datetime.strftime/strptime-compliant date format strings sorted by error rate
-
-  Examples:
-    >>> guess_bulk(['12-11-2014', '12-25-2014'])
-    ['%m-%d-%Y']
-    >>> guess_bulk(['12-11-2014', '25-25-2014'])
-    []
-    >>> guess_bulk(['12-11-2013', '13-8-2013', '05-25-2013', '12-25-2013'], error_rate=0.5)
-    ['%m-%d-%Y']
-  """
-  if error_rate == 0.0:
-    patterns = None
-    for date in dates:
-      guesses_patterns = guess(date)
-      if patterns is None:
-        patterns = guesses_patterns
-      else:
-        patterns = patterns.intersection(guesses_patterns)
-      if not patterns:
-        break   # No need to iterate more if zero patterns found
-    return list(patterns)
-  else:
-    found_dates = 0
-    pattern_counters = defaultdict(lambda: 0)
-    num_dates = len(dates)
-    min_num_dates_to_be_found = num_dates - num_dates * error_rate
-
-    for idx, date in enumerate(dates):
-      patterns = guess(date)
-      if patterns:
-        found_dates += 1
-      for pattern in patterns:
-        pattern_counters[pattern] = pattern_counters[pattern] + 1
-
-      # Early return if number of strings that can't be date is already over error rate
-      cells_left = num_dates - idx - 1
-      cannot_be_found = float(found_dates + cells_left) < min_num_dates_to_be_found
-      if cannot_be_found:
-        return []
-
-    patterns = [(v, k) for k, v in pattern_counters.items()
-                if v > min_num_dates_to_be_found]
-    patterns.sort(reverse=True)
-    return [k for (v, k) in patterns]
diff --git a/sandbox/grist/imports/test_dateguess.py b/sandbox/grist/imports/test_dateguess.py
deleted file mode 100644
index 8e960cd9..00000000
--- a/sandbox/grist/imports/test_dateguess.py
+++ /dev/null
@@ -1,102 +0,0 @@
-import unittest
-from imports.dateguess import guess, guess_bulk
-
-
-class TestGuesser(unittest.TestCase):
-  def assertDate(self, input_str, fmt_list):
-    guessed = guess(input_str)
-    self.assertEqual(set(guessed), set(fmt_list))
-
-  def assertDates(self, input_lst, error_rate, fmt_list):
-    guessed = guess_bulk(input_lst, error_rate=error_rate)
-    self.assertEqual(set(guessed), set(fmt_list))
-
-  def test_guess_dates(self):
-    self.assertDate('', [])
-    self.assertDate("2013-13-13", [])
-    self.assertDate("25/25/1911", [])
-
-    self.assertDate("2014-01-11", ['%Y-%m-%d', '%Y-%d-%m'])
-    self.assertDate("2014-11-01", ['%Y-%m-%d', '%Y-%d-%m'])
-    self.assertDate("1990-05-05", ['%Y-%m-%d', '%Y-%d-%m'])
-    self.assertDate("2013-12-13", ['%Y-%m-%d'])
-
-    self.assertDate("12/31/1999", ['%m/%d/%Y'])
-    self.assertDate("11/11/1911", ['%m/%d/%Y', '%d/%m/%Y'])
-    self.assertDate("5/9/1981", ['%m/%d/%Y', '%d/%m/%Y'])
-    self.assertDate("6/3/1985", ['%m/%d/%Y', '%d/%m/%Y'])
-
-    self.assertDate("12/31/99", ['%m/%d/%y'])
-    self.assertDate("11/11/11", ['%y/%m/%d', '%y/%d/%m', '%m/%d/%y', '%d/%m/%y'])
-    self.assertDate("5/9/81", ['%m/%d/%y', '%d/%m/%y'])
-    self.assertDate("6/3/85", ['%m/%d/%y', '%d/%m/%y'])
-
-    self.assertDate("31.12.91", ['%d.%m.%y'])
-    self.assertDate("4.4.87", ['%m.%d.%y', '%d.%m.%y'])
-
-    self.assertDate("13.2.8", ['%y.%m.%d', '%y.%d.%m'])
-    self.assertDate("31.12.1991", ['%d.%m.%Y'])
-    self.assertDate("4.4.1987", ['%m.%d.%Y', '%d.%m.%Y'])
-    self.assertDate("13.2.2008", ['%d.%m.%Y'])
-    self.assertDate("31.12.91", ['%d.%m.%y'])
-    self.assertDate("4.4.87", ['%m.%d.%y', '%d.%m.%y'])
-    self.assertDate("13.2.8", ['%y.%m.%d', '%y.%d.%m'])
-
-    self.assertDate("9 May 1981", ['%d %b %Y', '%d %B %Y'])
-    self.assertDate("31 Dec 1999", ['%d %b %Y'])
-    self.assertDate("1 Jan 2012", ['%d %b %Y'])
-    self.assertDate("3 August 2009", ['%d %B %Y'])
-    self.assertDate("2 May 1980", ['%d %B %Y', '%d %b %Y'])
-
-    self.assertDate("13/1/2012", ['%d/%m/%Y'])
-
-    self.assertDate("Aug 1st 2014", ['%b %dst %Y'])
-    self.assertDate("12/22/2015 00:00:00.10", ['%m/%d/%Y %H:%M:%S.%f'])
-
-  def test_guess_datetimes(self):
-    self.assertDate("Thu Sep 25 10:36:28 2003", ['%a %b %d %H:%M:%S %Y'])
-    self.assertDate("Thu Sep 25 2003 10:36:28", ['%a %b %d %Y %H:%M:%S'])
-    self.assertDate("10:36:28 Thu Sep 25 2003", ['%H:%M:%S %a %b %d %Y'])
-
-    self.assertDate("2014-01-11T12:21:05", ['%Y-%m-%dT%H:%M:%S', '%Y-%d-%mT%H:%M:%S'])
-    self.assertDate("2015-02-16T16:05:31", ['%Y-%m-%dT%H:%M:%S'])
-    # TODO remove all except first one
-    self.assertDate("2015-02-16T16:05", ['%Y-%m-%dT%H:%M', '%Y-%H-%MT%d:%m',
-                                         '%Y-%m-%HT%M:%d', '%Y-%d-%HT%M:%m'])
-    self.assertDate("2015-02-16T16", ['%Y-%m-%dT%H', '%Y-%m-%HT%d'])    #TODO remove second one
-
-    self.assertDate("Mon Jan 13 9:52:52 am MST 2014", ['%a %b %d %I:%M:%S %p %Z %Y'])
-    self.assertDate("Tue Jan 21 3:30:00 PM EST 2014", ['%a %b %d %I:%M:%S %p %Z %Y'])
-    self.assertDate("Mon Jan 13 09:52:52 MST 2014", ['%a %b %d %H:%M:%S %Z %Y'])
-    self.assertDate("Tue Jan 21 15:30:00 EST 2014", ['%a %b %d %H:%M:%S %Z %Y'])
-    self.assertDate("Mon Jan 13 9:52 am MST 2014", ['%a %b %d %I:%M %p %Z %Y'])
-    self.assertDate("Tue Jan 21 3:30 PM EST 2014", ['%a %b %d %I:%M %p %Z %Y'])
-
-    self.assertDate("2014-01-11T12:21:05", ['%Y-%m-%dT%H:%M:%S', '%Y-%d-%mT%H:%M:%S'])
-    self.assertDate("2015-02-16T16:05:31", ['%Y-%m-%dT%H:%M:%S'])
-    self.assertDate("Thu Sep 25 10:36:28 2003", ['%a %b %d %H:%M:%S %Y'])
-    self.assertDate("10:36:28 Thu Sep 25 2003", ['%H:%M:%S %a %b %d %Y'])
-
-    self.assertDate("2014-01-11T12:21:05+0000", ['%Y-%d-%mT%H:%M:%S%z', '%Y-%m-%dT%H:%M:%S%z'])
-    self.assertDate("2015-02-16T16:05:31-0400", ['%Y-%m-%dT%H:%M:%S%z'])
-    self.assertDate("Thu, 25 Sep 2003 10:49:41 -0300", ['%a, %d %b %Y %H:%M:%S %z'])
-    self.assertDate("Thu, 25 Sep 2003 10:49:41 +0300", ['%a, %d %b %Y %H:%M:%S %z'])
-
-    self.assertDate("2003-09-25T10:49:41", ['%Y-%m-%dT%H:%M:%S'])
-    self.assertDate("2003-09-25T10:49", ['%Y-%m-%dT%H:%M'])
-
-  def test_guess_bulk_dates(self):
-    self.assertDates(["11/11/1911", "25/11/1911", "11/11/1911", "11/11/1911"], 0.0, ['%d/%m/%Y'])
-    self.assertDates(["25/11/1911", "25/25/1911", "11/11/1911", "11/11/1911"], 0.0, [])
-    self.assertDates(["25/11/1911", "25/25/1911", "11/11/1911", "11/11/1911"], 0.5, ['%d/%m/%Y'])
-
-    self.assertDates(["25/11/1911", "25/25/1911", "11/11/1911", "11/11/1911"], 0.1, [])
-    self.assertDates(["23/11/1911", '2004 May 12', "11/11/1911", "11/11/1911"], 0.5, ['%d/%m/%Y'])
-
-    self.assertDates(['2004 May 12', "11/11/1911", "11/11/1911", "23/11/1911"], 0.5, ['%d/%m/%Y'])
-    self.assertDates(['2004 May 12', "11/11/1911", "11/11/1911", "23/11/1911"], 0.0, [])
-    self.assertDates(['12/22/2015', "12/22/2015 1:15pm", "2018-02-27 16:08:39 +0000"], 0.1, [])
-
-
-if __name__ == "__main__":
-  unittest.main()
diff --git a/sandbox/grist/imports/test_import_csv.py b/sandbox/grist/imports/test_import_csv.py
index 4a45513d..67364c8d 100644
--- a/sandbox/grist/imports/test_import_csv.py
+++ b/sandbox/grist/imports/test_import_csv.py
@@ -4,8 +4,6 @@ import textwrap
 import unittest
 from six import BytesIO, text_type
 import csv
-import calendar
-import datetime
 
 from imports import import_csv
 
@@ -22,9 +20,15 @@ def bytes_io_from_str(string):
 
 class TestImportCSV(unittest.TestCase):
 
-  def _check_col(self, sheet, index, name, typename, values):
+  def _check_col(self, sheet, index, name, _typename, values):
     self.assertEqual(sheet["column_metadata"][index]["id"], name)
-    self.assertEqual(sheet["column_metadata"][index]["type"], typename)
+    # Previously, strings were parsed and types were guessed in CSV imports.
+    # Now all data is kept as strings and the column type is left as Any
+    # so that type guessing and parsing can happen elsewhere.
+    # To avoid updating 85 calls to _check_col, the typename argument was kept but can be ignored,
+    # and all values are converted back to strings for comparison.
+    self.assertEqual(sheet["column_metadata"][index]["type"], "Any")
+    values = [text_type(v) for v in values]
     self.assertEqual(sheet["table_data"][index], values)
 
   def _check_num_cols(self, sheet, exp_cols):
@@ -40,18 +44,16 @@ class TestImportCSV(unittest.TestCase):
     self._check_col(sheet, 1, "int2", "Int", [5, '', ''])
     self._check_col(sheet, 2, "textint", "Text", ["12345678902345689", '', ''])
     self._check_col(sheet, 3, "bigint", "Text", ["320150170634561830", '', ''])
-    self._check_col(sheet, 4, "num2", "Numeric", [123456789.123456, '', ''])
-    self._check_col(sheet, 5, "bignum", "Numeric", [7.22597e+86, '', ''])
+    self._check_col(sheet, 4, "num2", "Numeric", ['123456789.1234560000', '', ''])
+    self._check_col(sheet, 5, "bignum", "Numeric", ['7.22597E+86', '', ''])
     self._check_col(sheet, 6, "date1", "DateTime",
-                    [calendar.timegm(datetime.datetime(2015, 12, 22, 11, 59, 00).timetuple()), None, None])
+                    [u'12/22/15 11:59 AM', u'', u''])
     self._check_col(sheet, 7, "date2", "Date",
-                    [calendar.timegm(datetime.datetime(2015, 12, 20, 0, 0, 0).timetuple()), None, None])
+                    [u'December 20, 2015', u'', u''])
     self._check_col(sheet, 8, "datetext", "Date",
-                    [calendar.timegm(datetime.date(2015, 12, 22).timetuple()), None, None])
+                    [u'12/22/2015', u'', u''])
     self._check_col(sheet, 9, "datetimetext", "DateTime",
-                    [calendar.timegm(datetime.datetime(2015, 12, 22, 0, 0, 0).timetuple()),
-                     calendar.timegm(datetime.datetime(2015, 12, 22, 13, 15, 0).timetuple()),
-                     calendar.timegm(datetime.datetime(2018, 2, 27, 16, 8, 39).timetuple())])
+                    [u'12/22/2015 00:00:00', u'12/22/2015 13:15:00', u'02/27/2018 16:08:39'])
 
 
   def test_user_parse_options(self):
@@ -68,7 +70,11 @@ class TestImportCSV(unittest.TestCase):
     self._check_col(parsed_file, 2, "PHONE", "Text", ['201-343-3434', '201.343.3434',
                                                       '2013433434', '(201)343-3434'])
     self._check_col(parsed_file, 3, "VALUE", "Int", [45, 4545, 0, 4])
-    self._check_col(parsed_file, 4, "DATE", "DateTime", [1519747719.0, 1519744119.0, 1519751319.0, None])
+    self._check_col(parsed_file, 4, "DATE", "DateTime",
+                    [u'2018-02-27 16:08:39 +0000',
+                     u'2018-02-27 16:08:39 +0100',
+                     u'2018-02-27 16:08:39 -0100',
+                     u''])
 
   def test_wrong_cols1(self):
     file_obj = bytes_io_from_str(textwrap.dedent(
diff --git a/sandbox/grist/imports/test_import_xls.py b/sandbox/grist/imports/test_import_xls.py
index d5a4fbc0..7a4a3326 100644
--- a/sandbox/grist/imports/test_import_xls.py
+++ b/sandbox/grist/imports/test_import_xls.py
@@ -16,31 +16,33 @@ class TestImportXLS(unittest.TestCase):
   def _check_col(self, sheet, index, name, typename, values):
     self.assertEqual(sheet["column_metadata"][index]["id"], name)
     self.assertEqual(sheet["column_metadata"][index]["type"], typename)
+    if typename == "Any":
+      # Convert values to strings to reduce changes to tests after imports were overhauled.
+      values = [str(v) for v in values]
     self.assertEqual(sheet["table_data"][index], values)
 
   def test_excel(self):
     parsed_file = import_xls.parse_file(*_get_fixture('test_excel.xlsx'))
 
-    # check that column type was correctly set to int and values are properly parsed
-    self.assertEqual(parsed_file[1][0]["column_metadata"][0], {"type": "Int", "id": "numbers"})
+    # check that column type was correctly set to numeric and values are properly parsed
+    self.assertEqual(parsed_file[1][0]["column_metadata"][0], {"type": "Numeric", "id": "numbers"})
     self.assertEqual(parsed_file[1][0]["table_data"][0], [1, 2, 3, 4, 5, 6, 7, 8])
 
     # check that column type was correctly set to text and values are properly parsed
-    self.assertEqual(parsed_file[1][0]["column_metadata"][1], {"type": "Text", "id": "letters"})
+    self.assertEqual(parsed_file[1][0]["column_metadata"][1], {"type": "Any", "id": "letters"})
     self.assertEqual(parsed_file[1][0]["table_data"][1],
       ["a", "b", "c", "d", "e", "f", "g", "h"])
 
-    # messy tables does not support bool types yet, it classifies them as ints
-    self.assertEqual(parsed_file[1][0]["column_metadata"][2], {"type": "Bool", "id": "boolean"})
-    self.assertEqual(parsed_file[1][False]["table_data"][2],
-      [True, False, True, False, True, False, True, False])
+    # 0s and 1s become Numeric, not boolean like in the past
+    self.assertEqual(parsed_file[1][0]["column_metadata"][2], {"type": "Numeric", "id": "boolean"})
+    self.assertEqual(parsed_file[1][0]["table_data"][2], [1, 0, 1, 0, 1, 0, 1, 0])
 
     # check that column type was correctly set to text and values are properly parsed
     self.assertEqual(parsed_file[1][0]["column_metadata"][3],
-                     {"type": "Text", "id": "corner-cases"})
+                     {"type": "Any", "id": "corner-cases"})
     self.assertEqual(parsed_file[1][0]["table_data"][3],
       # The type is detected as text, so all values should be text.
-      [u'=function()', '3.0', u'two spaces after  ',
+      [u'=function()', u'3.0', u'two spaces after  ',
         u'  two spaces before', u'!@#$', u'€€€', u'√∫abc$$', u'line\nbreak'])
 
     # check that multiple tables are created when there are multiple sheets in a document
@@ -51,23 +53,19 @@ class TestImportXLS(unittest.TestCase):
   def test_excel_types(self):
     parsed_file = import_xls.parse_file(*_get_fixture('test_excel_types.xlsx'))
     sheet = parsed_file[1][0]
-    self._check_col(sheet, 0, "int1", "Int", [-1234123, '', ''])
-    self._check_col(sheet, 1, "int2", "Int", [5, '', ''])
-    self._check_col(sheet, 2, "textint", "Text", ["12345678902345689", '', ''])
-    self._check_col(sheet, 3, "bigint", "Text", ["320150170634561830", '', ''])
+    self._check_col(sheet, 0, "int1", "Numeric", [-1234123, '', ''])
+    self._check_col(sheet, 1, "int2", "Numeric", [5, '', ''])
+    self._check_col(sheet, 2, "textint", "Any", ["12345678902345689", '', ''])
+    self._check_col(sheet, 3, "bigint", "Any", ["320150170634561830", '', ''])
     self._check_col(sheet, 4, "num2", "Numeric", [123456789.123456, '', ''])
     self._check_col(sheet, 5, "bignum", "Numeric", [math.exp(200), '', ''])
     self._check_col(sheet, 6, "date1", "DateTime",
              [calendar.timegm(datetime.datetime(2015, 12, 22, 11, 59, 00).timetuple()), None, None])
     self._check_col(sheet, 7, "date2", "Date",
              [calendar.timegm(datetime.datetime(2015, 12, 20, 0, 0, 0).timetuple()), None, None])
-    self._check_col(sheet, 8, "datetext", "Date",
-             [calendar.timegm(datetime.date(2015, 12, 22).timetuple()), None, None])
-    # TODO: all dates have different format
-    # self._check_col(sheet, 9, "datetimetext", "DateTime",
-    #          [calendar.timegm(datetime.datetime(2015, 12, 22, 0, 0, 0).timetuple()),
-    #           calendar.timegm(datetime.datetime(2015, 12, 22, 13, 15, 0).timetuple()),
-    #           calendar.timegm(datetime.datetime(2018, 02, 27, 16, 8, 39).timetuple())])
+    self._check_col(sheet, 8, "datetext", "Any", ['12/22/2015', '', ''])
+    self._check_col(sheet, 9, "datetimetext", "Any",
+                    [u'12/22/2015', u'12/22/2015 1:15pm', u'2018-02-27 16:08:39 +0000'])
 
   def test_excel_type_detection(self):
     # This tests goes over the second sheet of the fixture doc, which has multiple rows that try
@@ -81,23 +79,20 @@ class TestImportXLS(unittest.TestCase):
                      1454544000.0, 1199577600.0, 1451692800.0, 1451549340.0, 1483214940.0])
     self._check_col(sheet, 1, "float_not_int", "Numeric",
                     [1,2,3,4,5,"",6,7,8,9,10,10.25,11,12,13,14,15,16,17,18])
-    self._check_col(sheet, 2, "int_not_bool", "Int",
+    self._check_col(sheet, 2, "int_not_bool", "Any",
                     [0, 0, 1, 0, 1, 0, 0, 1, 0, 2, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0])
-    self._check_col(sheet, 3, "float_not_bool", "Numeric",
+    self._check_col(sheet, 3, "float_not_bool", "Any",
                     [0, 0, 1, 0, 1, 0, 0, 1, 0, 0.5, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0])
-    self._check_col(sheet, 4, "text_as_bool", "Bool",
+    self._check_col(sheet, 4, "text_as_bool", "Any",
                     [0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0])
-    self._check_col(sheet, 5, "int_as_bool", "Bool",
+    self._check_col(sheet, 5, "int_as_bool", "Numeric",
                     [0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0])
-    self._check_col(sheet, 6, "float_not_date", "Numeric",
+    self._check_col(sheet, 6, "float_not_date", "Any",
                     [4.0, 6.0, 4.0, 4.0, 6.0, 4.0, '--', 6.0, 4.0, 4.0, 4.0, 4.0, 4.0, 6.0, 6.0,
                      4.0, 6.0, '3-4', 4.0, 6.5])
     self._check_col(sheet, 7, "float_not_text", "Numeric",
-                    [-10.25, -8.00, -5.75, -3.50, "n/a", 1.00, "   ???   ", 5.50, "", "-",
+                    [-10.25, -8.00, -5.75, -3.50, "n/a", '  1.  ', "   ???   ", 5.50, "", "-",
                      12.25, 0.00, "", 0.00, "--", 23.50, "NA", 28.00, 30.25, 32.50])
-    self._check_col(sheet, 8, "dollar_amts", "Numeric",
-                    [0.00, 0.75, 1.50, '', 3.00, 0.00, 0.75, 1.50, '--', 3.00, 1234.56, 1000,
-                     1001.50, '-', 3000000.000, 0000.00, 1234.56, 1000, 1001.50, 1000.01])
 
   def test_excel_single_merged_cell(self):
     # An older version of xlrd had a bug where a single cell marked as 'merged' would cause an
@@ -107,11 +102,11 @@ class TestImportXLS(unittest.TestCase):
     self.assertEqual(tables, [{
       'table_name': u'Transaction Report',
       'column_metadata': [
-        {'type': 'Text', 'id': u''},
+        {'type': 'Any', 'id': u''},
         {'type': 'Numeric', 'id': u'Start'},
         {'type': 'Numeric', 'id': u''},
         {'type': 'Numeric', 'id': u''},
-        {'type': 'Text', 'id': u'Seek no easy ways'},
+        {'type': 'Any', 'id': u'Seek no easy ways'},
       ],
       'table_data': [
         [u'SINGLE MERGED', u'The End'],
@@ -133,15 +128,15 @@ class TestImportXLS(unittest.TestCase):
     self.assertEqual(tables, [{
       'table_name': u'Sheet1',
       'column_metadata': [
-        {'id': 'a', 'type': 'Text'},
+        {'id': 'a', 'type': 'Any'},
         {'id': 'b', 'type': 'Date'},
-        {'id': 'c', 'type': 'Text'},
-        {'id': 'd', 'type': 'Text'},
+        {'id': 'c', 'type': 'Any'},
+        {'id': 'd', 'type': 'Any'},
         {'id': 'e', 'type': 'Numeric'},
-        {'id': 'f', 'type': 'Int'},
-        {'id': 'g', 'type': 'Date'},
+        {'id': 'f', 'type': 'Numeric'},
+        {'id': 'g', 'type': 'Any'},
         {'id': 'h', 'type': 'Date'},
-        {'id': 'i', 'type': 'Bool'},
+        {'id': 'i', 'type': 'Numeric'},
       ],
       'table_data': [
         [u'21:14:00'],
@@ -150,9 +145,9 @@ class TestImportXLS(unittest.TestCase):
         [u'10:20:30'],
         [4.180902777777778],
         [20],
-        [-6106060800.0],
+        [u'7/4/1776'],
         [205286400.0],
-        [False],    # This is not great either, we should be able to distinguish 0 from FALSE.
+        [0],
       ],
     }])
 
diff --git a/sandbox/grist/parse_data.py b/sandbox/grist/parse_data.py
index 460fa6f8..c5aa1cf4 100644
--- a/sandbox/grist/parse_data.py
+++ b/sandbox/grist/parse_data.py
@@ -7,13 +7,11 @@ dictionary with "type" and "data" fields, where "type" is a Grist type string, a
 of values. All "data" lists will have the same length.
 """
 
-from imports import dateguess
 import datetime
 import logging
 import re
 import messytables
 import moment # TODO grist internal libraries might not be available to plugins in the future.
-import dateutil.parser as date_parser
 import six
 from six.moves import zip, xrange
 
@@ -25,12 +23,17 @@ log = logging.getLogger(__name__)
 
 
 # Our approach to type detection is different from that of messytables.
-# We first go through each cell in a sample of rows, trying to convert it to each of the basic
+# We first go through each cell in a sample of rows, checking if it's one of the basic
 # types, and keep a count of successes for each. We use the counts to decide the basic types (e.g.
 # numeric vs text). Then we go through the full data set converting to the chosen basic type.
 # During this process, we keep counts of suitable Grist types to consider (e.g. Int vs Numeric).
 # We use those counts to produce the selected Grist type at the end.
 
+# Previously string values were used here for type guessing and were parsed to typed values.
+# That process now happens elsewhere, and this module only handles the case
+# where the imported data already contains actual numbers or dates.
+# This happens for Excel sheets but not CSV files.
+
 
 class BaseConverter(object):
   @classmethod
@@ -57,50 +60,19 @@ class BaseConverter(object):
 
 
 class NumericConverter(BaseConverter):
-  """Handles numeric values, including Grist types Numeric and Int."""
-
-  # A number matching this is probably an identifier of some sort. Converting it to a float will
-  # lose precision, so it's better not to consider it numeric.
-  _unlikely_float = re.compile(r'\d{17}|^0\d')
-
-  # Integers outside this range will be represented as floats. This is the limit for values that can
-  # be stored in a JS Int32Array.
-  _max_js_int = 1<<31
-
-  # The thousands separator. It should be locale-specific, but we don't currently have a way to
-  # detect locale from the data. (Also, the sandbox's locale module isn't fully functional.)
-  _thousands_sep = ','
+  """Handles the Grist Numeric type"""
 
   @classmethod
   def convert(cls, value):
     if type(value) in six.integer_types + (float, complex):
       return value
-    if type(value) in (str, six.text_type) and not cls._unlikely_float.search(value):
-      return float(value.strip().lstrip('$').replace(cls._thousands_sep, ""))
     raise ValueError()
 
-  @classmethod
-  def _is_integer(cls, value):
-    ttype = type(value)
-    if ttype == int or (ttype == float and value.is_integer()):
-      return -cls._max_js_int <= value < cls._max_js_int
-    return False
-
   @classmethod
   def get_grist_column(cls, values):
-    if all(cls._is_integer(v) for v in values):
-      return ("Int", [int(v) for v in values])
     return ("Numeric", values)
 
 
-class DateParserInfo(date_parser.parserinfo):
-  def validate(self, res):
-    # Avoid this bogus combination which accepts plain numbers.
-    if res.day and not res.month:
-      return False
-    return super(DateParserInfo, self).validate(res)
-
-
 class SimpleDateTimeConverter(BaseConverter):
   """Handles Date and DateTime values which are already instances of datetime.datetime."""
 
@@ -124,66 +96,18 @@ class SimpleDateTimeConverter(BaseConverter):
     return grist_type, grist_values
 
 
-class DateTimeCoverter(BaseConverter):
-  """Handles dateformats by guessed format."""
-
-  def __init__(self, date_format):
-    self._format = date_format
-
-  def convert(self, value):
-    if value == "":
-      return None
-    if type(value) in (str, six.text_type):
-      # datetime.strptime doesn't handle %z and %Z tags in Python 2.
-      if '%z' in self._format or '%Z' in self._format:
-        return date_parser.parse(value)
-      else:
-        try:
-          return datetime.datetime.strptime(value, self._format)
-        except ValueError:
-          return date_parser.parse(value)
-
-    raise ValueError()
-
-  def _is_date(self, value):
-    return value is None or value.time() == datetime.time()
-
-  def get_grist_column(self, values):
-    grist_type = "Date" if all(self._is_date(v) for v in values) else "DateTime"
-    grist_values = [(v if (v is None) else moment.dt_to_ts(v))
-                    for v in values]
-    return grist_type, grist_values
-
-
-class BoolConverter(BaseConverter):
-  """Handles Boolean type."""
-
-  _true_values = (1, '1', 'true', 'yes')
-  _false_values = (0, '0', 'false', 'no')
-
-  @classmethod
-  def convert(cls, value):
-    v = value.strip().lower() if type(value) in (str, six.text_type) else value
-    if v in cls._true_values:
-      return True
-    elif v in cls._false_values:
-      return False
-    raise ValueError()
-
-  @classmethod
-  def get_grist_column(cls, values):
-    return ("Bool", values)
-
-
-class TextConverter(BaseConverter):
-  """Fallback converter that converts everything to strings."""
+class AnyConverter(BaseConverter):
+  """
+  Fallback converter that converts everything to strings.
+  Type guessing and parsing of the strings will happen elsewhere.
+  """
   @classmethod
   def convert(cls, value):
     return six.text_type(value)
 
   @classmethod
   def get_grist_column(cls, values):
-    return ("Text", values)
+    return ("Any", values)
 
 
 class ColumnDetector(object):
@@ -194,7 +118,7 @@ class ColumnDetector(object):
   """
   # Converters are listed in the order of preference, which is only used if two converters succeed
   # on the same exact number of values. Text is always a fallback.
-  converters = [SimpleDateTimeConverter, BoolConverter, NumericConverter]
+  converters = [SimpleDateTimeConverter, NumericConverter]
 
   # If this many non-junk values or more can't be converted, fall back to text.
   _text_threshold = 0.10
@@ -221,19 +145,11 @@ class ColumnDetector(object):
         self._counts[i] += 1
 
   def get_converter(self):
-    if sum(self._counts) == 0:
-      # if not already guessed as int, bool or datetime then we should try to guess date pattern
-      str_data = [d for d in self._data if isinstance(d, six.string_types)]
-      data_formats = dateguess.guess_bulk(str_data, error_rate=self._text_threshold)
-      data_format = data_formats[0] if data_formats else None
-      if data_format:
-        return DateTimeCoverter(data_format)
-
     # We find the max by count, and secondarily by minimum index in the converters list.
     count, neg_index = max((c, -i) for (i, c) in enumerate(self._counts))
     if count > 0 and count >= self._count_nonjunk * (1 - self._text_threshold):
       return self.converters[-neg_index]
-    return TextConverter
+    return AnyConverter
 
 
 def _guess_basic_types(rows, num_columns):